Article 7440 of comp.org.decus:
Path: nntpd.lkg.dec.com!crl.dec.com!crl.dec.com!bloom-beacon.mit.edu!spool.mu.edu!usenet.eel.ufl.edu!news.mathworks.com!uunet!in1.uu.net!eisner!mcmahon_b
From: mcmahon_b@eisner.decus.org (Brian McMahon, Info-VAX Refugee)
Newsgroups: comp.org.decus,vmsnet.decus.journal
Subject: DECUServe Journal September 1995
Message-ID: <1995Sep26.220803.11367@eisner>
Date: 26 Sep 95 22:08:03 -0400
Organization: DECUServe
Approved: mcmahon_b@eisner.decus.org
Lines: 3654
Xref: nntpd.lkg.dec.com comp.org.decus:7440 vmsnet.decus.journal:11
The DECUServe Journal
---------------------
September, 1995
From the Editors' Keyboard . . . . . . . . . . . . . 2
What's Inside, Words to and from the Weary
Large and Small Web Graphics . . . . . . . . . . . . 3
How Big is Too Big for On-line Browsing?
Stripe/Shadow and Cacheing . . . . . . . . . . . . . 6
Do Cacheing Products Help with RAID 0+1?
TeamLinks SDK . . . . . . . . . . . . . . . . . . . 8
Software Development Kit for TeamLinks
Aliases for ALL-IN-1 Account Names . . . . . . . . 15
On Shortening Existing Usernames, and DDS
PMDF and Message Router . . . . . . . . . . . . . 17
Experiences with PMDF V5.0 and ALL-IN-1
Reverse Linking of Shareable Images . . . . . . . 20
Implementing (or Avoiding) Circular Linking
DCPS Memory Leak Fixed . . . . . . . . . . . . . . 23
No More Memory Leak in DECprint Supervisor V1.2
800 bpi Tape Drive Needed . . . . . . . . . . . . 24
Remember Those? What If You Need One Now?
Video for Alpha 2000-300 . . . . . . . . . . . . . 26
What Video Can the 2000-300 Handle?
RAID Options on OpenVMS . . . . . . . . . . . . . 30
Deciding Which RAID Is Best for You
UAS/UAF Merge . . . . . . . . . . . . . . . . . . 33
Merging PATHWORKS UAS and OpenVMS UAF under NT
DHU-11 Installation Problem . . . . . . . . . . . 35
Trouble Adding a Board to a PDP-11/84
UNIX and DOS on the Same Machine . . . . . . . . . 37
Or, "How to Have Your Linux and DOS, Too"
POLYCENTER Console Manager . . . . . . . . . . . . 40
PCM Experiences, Testimonials, and Tips
Slow System Startup . . . . . . . . . . . . . . . 45
OpenVMS Startup Slowed by INSTALL
OpenVMS System with Expired Licenses . . . . . . . 47
Or, How to Crack OpenVMS from the Console
Windows NT or Windows 95 Clients? . . . . . . . . 49
What to Develop for?
The DECUServe Journal September, 1995 Page 2
From the Editors' Keyboard
From the Editors' Keyboard
---- --- -------- --------
You might think of this issue as a buzzword grab-bag of sorts. A
quick scan of the table of contents will reveal several articles
relating to one or more "hot topics" in the computing industry. We
have Web stuff, RAID, TeamLinks, Windows NT, e-mail integration, and
finish up with a long discussion of whether one should target
development for Windows 95 clients or not.
But what, you might say, if you don't care for the cutting
(bleeding) edge? Suppose glossy brochures and jargon-laden sales
pitches give you a rash? Fear not -- the DECUServe subscribers (and
your long-laboring editors) haven't forgotten the old reliables.
Here, too, you can read about a problem with a hardware addition to
a PDP-11, and a subscriber's quest for a working 800 bpi tape drive.
In short, this issue's contents are about as varied as Iowa weather.
We hope you enjoy the variety more than the farmers have been
recently!
* * * * *
The DECUServe Journal September, 1995 Page 3
Large and Small Web Graphics
Large and Small Web Graphics
----- --- ----- --- --------
Abstract:
When are graphic images in a Web page too big? When are they too
small? As you might imagine, there is no simple answer. The
following notes contain arguments in favor of "big" as well as
"small," and your mileage is almost certain to vary.
Participants: Gus Altobello,
Charlie Byrne, Bob Hassinger, J.M. Ivler, Bart Lederman, Benjamin
Pena.
Conference: WWW
Note 51.0, 31-Jul-1995
Lederman: IMG: What size pictures do you want to browse?
--------------------------------------------------------
When you're browsing around for WWW stuff, how big can / should an
image be?
I realize this depends a lot on what's being shown, but I'm looking for
general guidelines for limits. For example, if you've run across
images that were too small to really tell you anything, or images which
are too big and take too long to view. (This is for images in general
WWW pages, not where there is a pointer that says "you can get a
picture of this but it's xxx in size.) Also, this is for pictures of
'real' stuff, not drawings or symbols which can be made relatively
small and still be recognizable.
Note 51.1, 31-Jul-1995
Byrne: depends...
-----------------
At midnight, 100k is fine.
At noon, 1 byte is too big.
(Sorry, I know that is no help, but that's how I feel).
Don't forget resolution 75 .vs. 150 . vs. 300 etc
increases size geometrically not linearly.
Also consider multi-color .vs. grey scale.
The DECUServe Journal September, 1995 Page 4
Large and Small Web Graphics
Note 51.2, 1-Aug-1995
Lederman: I don't think colors or resolution counts here.
---------------------------------------------------------
All of the images we are considering, and all the one's I've seen, are
GIF files. This means a maximum of 256 colors, and it doesn't matter
if they're 'real' colors, or shades of grey. And I believe that GIFs
only specify the number of pixels vertically and horizontally: the
number of pixels per inch depends on the device on which it is viewed.
Note 51.3, 1-Aug-1995
Ivler: 1 meg of images loading over a 9600b line...
---------------------------------------------------
I am from the thumbnail school...
I think that a corporate logo is fine, as long as it doesn't take
up 80% of the page and loads in less than a day. Remember, it's
your corporate image that they get to see :-)
Hmmm, consider it this way... almost 70% of the people browsing
the web use readers that have connectivity at 9600-14.4. How long
are you willing to sit there and wait as an image loads at that
speed...
Thumbnail the image and then provide a link from the thumbnail to
a larger gif. This allows the viewer a chance to get a feel for
the image and see it (partially) before deciding to download the
entire image (it's also nice to include image size in there so
they can determine if they want to load a 276K image at 9600).
Of course, these are all personal opinions. I know of people that
think imagemaps are wonderful (yeah, they are cute and all, but
unless they can be kept "small and fast" (jpg) loading, they are a
bust, not a wonderous thing...). So, before you create a page
chockful of images and netscape enhancements (like complex
backgrounds) consider the speed of access and the viewer, you
know, they person that your trying to reach.
One last note... I have received more compliments on the "low
image" Magic Castle pages than requests for them to be "imaged".
In fact, only Castle members who are on committees (like
membership and marketing" seem to want to image enhance the
pages... most of the readers like the fact that they contain
"information" first...
Note 51.4, 1-Aug-1995
Altobello: ...can't be bothered...
----------------------------------
Is there a reason folks go with .gif files rather than the smaller
compressed .jpg files?
The DECUServe Journal September, 1995 Page 5
Large and Small Web Graphics
I see this argument all the time in various binaries groups (and in
some groups I suppose some folks would be more desperate to be able to
download more... :-) but the ".gifs are better than .jpgs" line of
argument makes no sense to me, based on my own comparisons.
Is it that .jpg viewers are rarer than .gif viewers?
FWIW, while I don't much care if little thumbnails are all over the
page, what really annoys me (and often causes me to abort and go to
another site) is some fool with several huge graphics on their
top-level page that take forever (>1min) to load at 14.4k. I will go
and look at such things (like the Nagasaki Remembered site, with over
20 photos), but only if I have a choice in the matter.
From my point of view, there's too much out there to tolerate poor
design. If the first page is set up poorly (weird fonts, strange
"features" required of the browser, huge (and often stupid) graphics)
there's a good chance the rest is more of the same, so why bother?
Note 51.5, 1-Aug-1995
Altobello: I don't believe in "too small"; others disbelieve "too big".
-----------------------------------------------------------------------
BTW, in answer to another question in the .0 post, I've been at sites
that had "real" images, paintings, as thumbnails. While I couldn't run
an Art Appreciation course on these, they were certainly clear enough
for me to identify the paintings I wanted to see more of, and clicking
on them brought up the "big one".
I can't say I've ever seen an image which was "too small", and have
seen FAR too many that were either "too big" or "not needed". "Not
needed" is the most annoying category.
Go read those books that come with the HP color inkjet printers, the
ones that say things like "use color sparingly", and apply it to
pictures. Remember what happened when folks first started using TeX
and produced memos and reports that looked like ransom notes, from the
excessive use of font styles.
KISS, y'know? Can't go much wrong there.
Note 51.6, 2-Aug-1995
Pena: an idea..
---------------
Although.. sometimes I get annoyed when my machine is hanging off a T-1
line, and there's a zillion of itty bitty graphics :) when I'm
perfectly capable of appreciating big/detailed images.
I know this might be more work, but might satisfy more people overall:
offer them two URL's: both with same text but different in
The DECUServe Journal September, 1995 Page 6
Large and Small Web Graphics
size/detail/resolution of images.
Communication is essentially trying to reach as many people as
possible.. That's just my perspective. Take it as you will. :)
Note 51.7, 3-Aug-1995
Hassinger: 64 by 128 - i.e. small for on-screen
-----------------------------------------------
There is no one answer of course - but I know you know that.
GIF vs JPEG - JPEG is nice because it gives high compression meaning
lower transmission times. However it does introduce artifacts and
degradation when you deal with high compression levels and higher
quality presentation means on the receiving end (e.g. that let you see
the degradation). Also the receiving end has a higher compute load to
decompress the image - on slower Macs I see the gain in communication
time lost in the time the Mac spends.
You manage GIF differently - reduce the number of colors to 40 or 50
for example - it helps the browsers, particularly if you have more than
one image on the browser screen that uses different color palettes, and
it makes the files smaller in general. You get the transparency option
and the interleaved, progressive display option in the bargain too.
I think the real bottom line right now is that more browsers handle
GIFs more easily and more gracefully. For on-screen, moderate size
images, go with GIF. For high quality, larger image download for other
uses than on-screen, JPEG may be best.
And test your pages against the typical target browsers you expect to
serve to, and try to do it over communications paths typical of what
will be used. Then make it work so you are satisfied. (e.g.
empirical evaluation of how long it takes...)
This heuristic (holistic?) approach seems to be the only really viable
way to answer these questions at this point in such a rapidly changing
field with so much anarchy out there.
Stripe/Shadow and Cacheing
------------- --- --------
Abstract:
The question in the next series of notes concerns whether or not
host-based cacheing products are beneficial in an environment
featuring large database files and stripe/shadow sets (a.k.a. RAID
01). The poster asked for "knowledgeable sources," and promptly
found them.
The DECUServe Journal September, 1995 Page 7
Stripe/Shadow and Cacheing
Participants: Larry Clegg,
Jerry Evans, Keith Parris.
Conference: 3RD_PARTY_VMS_SOFTWARE
Note 412.0, 18-Aug-1995
Evans: Caching Question
-----------------------
Dumb question time.
I need some discussion from knowledgeable sources concerning disk-caching
products in an environment (ALPHA 2100, VMS 6.1) where our primary
application resides on a RAID-0+1 shadowed stripeset equivalent.
The application is from EMS (Effective Management Systems, Milwaukee) and
is an integrated business system, with accounting, inventory, shop floor,
order processing, planning & etc Modules.
I have heard that caching products have little or no impact when you have
the stripe/shadow type setup for huuuge database-type files.
Comments, Please. Thanks.
Note 412.1, 18-Aug-1995
Clegg: Shadow Ok, Stripe Maybe
------------------------------
> Dumb question time.
No such thing!
> I have heard that caching products have little or no impact when you have
> the stripe/shadow type setup for huuuge database-type files.
I'm using PerfectCache from Raxco on my VAXen with shadowed disks. The
shadowing doesn't really play into caching performance since caching,
PerfectCache at least, is only caching read i/o's not the writes. I can see
where using striped sets would affect caching performance since one read would
have to be potentially satisfied by multi-spindle disk accesses. Unless the
caching software was caching across all the volumes and had all the required
data already in memory there wouldn't be a performance gain. Perhaps you
should contact some of the cacheing software vendors and pose these questions.
Note 412.2, 18-Aug-1995
Parris: Caching and RAID
------------------------
It's not clear from .0 whether you're doing the RAID and/or the caching in
the host or a controller.
Read caching (as host-based products provide) benefits performance in two ways:
The DECUServe Journal September, 1995 Page 8
Stripe/Shadow and Cacheing
1) when there is high locality of reference (re-reading the same stuff a lot
of the time)
2) when the product supports pre-fetching, and access is sequential
Read caching in controllers and in disks themselves has similar benefits. If a
controller provides safe (non-volatile) write-back cache (like the option for
the HSJ/HSD/HSZ series or the just-announced battery-backup option for the
KZPSC), it can also help significantly with performance on writes.
The performance effect of the RAID 0+1 array is somewhat orthogonal to the
benefits provided by caching.
1) the availability of multiple members from which to satisfy a read can
benefit read performance (but not all shadowing/mirroring/RAID-1 products
do this; the KZESC/KZPSC/SWXCR have a "master" member from which all reads
are done, while host-based volume shadowing selects members based on
local-vs-remote location and queue depth).
2) having to update both members of a shadow set on a write cuts performance a
little, unless the controller can hide this latency using write-back cache.
Note that read caching tends to reduce the read/write ratio, so you get
less of the benefit of (1) and more of the penalty of (2) when it is
present up at the host level.
3) the disk striping (RAID-0) placement of data tends to spread the I/O load
evenly across all the available spindles, resulting in lower average
response times. This works even for totally random requests, where
caching would be of little benefit.
Database products tend to make larger I/Os than typical OpenVMS applications,
so a larger chunk size may be appropriate for the RAID array.
Locality of reference is something that seems to be much harder to guess at or
measure than the I/O size or the read/write ratio, so it's a bit harder to
guess whether read caching will help. Write caching should definitely help the
write performance, since that works regardless of locality of reference.
Complicating all this is the fact that many database packages do their own
caching internally.
TeamLinks SDK
--------- ---
Abstract:
Where to get the TeamLinks Software Development Kit on the net,
followed by a review of the SDK, provided by DECUServe's very own
Don Vickers.
The DECUServe Journal September, 1995 Page 9
TeamLinks SDK
Participant: Don Vickers.
Conference: ALL-IN-1
Note 1143.0, 15-Jul-1995
Vickers: TeamLinks SDK on the Internet
--------------------------------------
The URL for the TeamLinks SDK is:
ftp://ftp.service.digital.com/teamlink/tlftp/win/tl25/tlsdk/tlsdk.exe
This is a rather long path and is subject to change so you may wish to
start browsing at:
ftp://ftp.service.digital.com/teamlink
There is another file in the directory which contains just the
documentation for the SDK. You do not need this unless you want to see
what is in the SDK. It is included in the tlsdk.exe.
The teamlink directory is the root for all the public TeamLinks files.
For example, the latest ECO (5) for TeamLinks V2.1 is there.
Note 1143.1, 7-Aug-1995
Vickers: TeamLinks V2.5 Software Development Kit - A review
-----------------------------------------------------------
The TeamLinks SDK (Software Development Kit) is an indispensable tool for
anyone implementing TeamLinks. Customizing and extending TeamLinks to best
conform to each organization's needs is important.
The SDK is a collection of software and documentation which includes:
* Sample working programs written in C and Visual Basic
* Definition files for C and Visual Basic
* Help Files
* Documentation covering the various aspects of the TeamLinks APIs
* Software licenses
The most complete document for programming TeamLinks is in "Working with
TeamLinks, 2nd Edition", by Tony Redmond (Digital Press, part number
EY-T080E-DP and ISBN: 1-55558-140-4). Tony Redmond is the 'editor' of the SDK
as well as author of the book. The book refers to many parts of the SDK as
examples. The book is almost as valuable as the SDK. My advice is to buy the
book from someplace else if you wish to save a bit of money. Purchasing the
book from Digital will cost a little more and might help us Digital
stockholders.
Constructive Criticisms
The SDK is a most powerful tool in implementing TeamLinks. As with anything
done by a human, it can be improved.
The DECUServe Journal September, 1995 Page 10
TeamLinks SDK
The text in the SDK claims a 'regular update' schedule. The definition of
'regular' may differ from country to country and person to person. Predicting
the times may require a crystal ball. The most recent SDK, dated June 1995,
still refers to V2.1 as the most recent version of TeamLinks in some key
documents. Having a really regular update schedule would be a nice
improvement. Clearly, there are finite resources in Digital for this sort of
thing but making the SDK more predictable would seem to have a good return on
their investment.
The primary SDK document contains a list of contents but there is no indication
of changes. Having a history or last update for each section of the SDK would
be a most welcome addition.
The SDK is not supported by Digital since the contents do not go through the
engineering processes. The contents of the SDK have differing levels of
completeness and quality. Most of the applications are solid but many also
have small 'features' in various customer environments.
The only current 'support' channels are through the DECPCI conference on
CompuServe or via your local Digital office, if any. Having a more universally
available support channel would be a good improvement. The SDK has recently
been placed on the Internet which is a nice step. Adding a Web site and mail
address would be a good step.
Obtaining the SDK
The SDK is free to anyone.
The TeamLinks SDK is available on CompuServe, in the DECPCI conference, Library
#6 (TeamLinks/Int Apps), TLSDK.EXE.
It is also available via FTP at the following URL (universal resource locator):
ftp://ftp.service.digital.com/teamlink/tlftp/win/tl25/tlsdk/tlsdk.exe
This URL may change in the future to reduce the number of sub directories.
Should this occur, you should be able to start your quest for the file at the
teamlink directory.
Installing the SDK
The TeamLinks SDK is available as a self-extracting DOS executable (TLSDK.EXE).
Place the TLSDK.EXE into a directory you will use as the base for the SDK.
TLSDK is a popular choice for the directory name.
At the DOS prompt invoke TLSDK.EXE using the /D qualifier to have the
sub-directories created. I have forgotten this D detail a few times, so far.
The top level directory will contain only the directories and the TLSDK.EXE
file unless there were other files in the base directory before the explosion.
If you are going to use the TeamLinks Visual Basic custom controls, you must
copy the TLDEVO.LIC from the \TEAMLINK directory into the \WINDOWS directory.
The DECUServe Journal September, 1995 Page 11
TeamLinks SDK
You should submit a signed copy of the PAPER.LIC license agreement to your
local Digital office if you are going to implement projects using these
controls so that Digital is aware of this.
Documentation
The documentation should be your first stop when starting to use the SDK. Most
of the documents are in Microsoft Word format but there are a few text files
and one in PostScript just for fun. If you are foolish enough to not being on
the "Let's help build the Gates mansion plan" and do not have Word, you can use
the TeamLinks converters to change the format.
The documentation describes the various APIs (Application Programming
Interfaces) used by the various components of TeamLinks. The TLSDK.DOC
document is the primary documentation for the SDK. It describes much of the
content of the SDK and replaces much of the information in the CUSTOM.WRI
document.
Below are the documents currently in the SDK:
TLSDK.DOC Overview of the SDK including language enhancements
AIDA.TXT ALL-IN-1 Data Set Server functions (AIDA.DLL)
FCS.TXT ALL-IN-1 File Cabinet Server functions.
CONN25.DOC TeamLinks Connection Manager (CONNECT.DLL)
FILECA25.DOC TeamLinks File Cabinet integration module (FILECAB.DLL)
CFCAPI21.DOC API for the TeamLinks Common File Cabinet (CFC.DLL)
ADDRES20.DOC Mail user address validation specification. (CFC API Supplement)
TLNOTIFY.DOC Documentation for TLNOTIFY.VBX.
PROFIL20.DOC Mail user profile access specification. (CFC API Supplement)
RESERV20.DOC File Cabinet reservation specification. (CFC API Supplement)
MAILDDE.DOC TeamLinks Messaging DDE Server
DDS.PS Distributed Directory Services (DDS) API. (DDSX400.DLL)
Supporting software
There are a number of modules on the SDK that provide support for building
applications that extend or enhance TeamLinks. These modules allow for
creating applications that look and feel like TeamLinks and use the various
services provided by TeamLinks.
Applications
There is range of applications provided on the SDK from complete and polished
applications to rough examples. Each provides the source code so that you can
see how the work is performed. This allows you to create your own applications
using the samples as a base.
There are over 30 complete and mostly operational applications in the SDK.
This review is already longer than good tastes and writing allow so we will not
discuss them all. I will mention a handful of the most powerful and useful
applications right out of the box.
The most common use of TeamLinks is for electronic mail and by far the most
The DECUServe Journal September, 1995 Page 12
TeamLinks SDK
common request from most sites is to improve the way mail notification is
performed. The TeamLinks engineers have done a nice job of improving the new
mail notification over the past few releases. However, many folks still prefer
to have a better view of the mail waiting their attention. Even folks who have
been using ALL-IN-1 like to have a count of new messages that is automatic.
There are two nice applications which provide added value in this area.
MAILCNT in the \AIDA sub-directory provides an appearance that emulates the
ALL-IN-1 forms that display the count of new mail and the person's name.
MAILCNT will update the TeamLinks Information Manager window and icon caption
text to show the count of new mail messages as well as the person's full name.
If TeamLinks is not running, MAILCNT creates the world's smallest window that
contains just the new mail count.
TLNWMAIL in the \MAIL sub-directory provides a separate window which allows the
person to display the new mail messages. The application allows for changing
the windows contents and to read mail messages directly from the window.
The IOSPrint application in the \IOSP sub-directory implements a means to allow
TeamLinks clients to print to the queued printers on the ALL-IN-1 server. This
can be most useful for sites migrating from ALL-IN-1 as it provides a means to
allow clients to use the shared printers in the same manner as before the
migration.
It seems that virtually any site currently running ALL-IN-1 and implementing
TeamLinks is concerned about being able to utilize the applications they have
in ALL-IN-1 in the TeamLinks environments. It is rather easy to provide a
TeamLinks version of some ALL-IN-1 applications which are implemented via FMS
forms and scripts.
The PHPENT application in the \AIDA sub-directory is a very nice example of how
easy it is map an ALL-IN-1 Entry form to a Visual Basic panel. The name of the
application is rather misleading. It is named after the ALL-IN-1 form, PHPENT,
that defines the personal telephone directory. I assume this was the original
form of interest when the application was first developed. The current version
of the PHPENT allows you to select any available ALL-IN-1 Entry form along with
the fields to be included. This can be used as the basis for developing 'real'
applications that can perform direct mapping of data defined by Entry forms.
The SUBSCRIP application also in the \AIDA sub-directory also suffers from a
misleading name. At least my simple mind keeps thinking that SUBSCRIP refers
to something relating to the Subscribers dataset or something related. In
fact, SUBSCRIP is a concatenation of Submit Script. SUBSCRIP allows for a
generalized means for submitting scripts to the ALL-IN-1 script symbiont from
the PC. It serves as an example of how one can trigger significant events on
the ALL-IN-1 server from PC clients.
The TLPM application in the \PMEMOS also suffers from a naming 'feature'. The
directory name is the same name as an old moderately popular ALL-IN-1 ASSETS
package that provided the ability to send telephone messages to people via
electronic mail. This application demonstrates a more complete and polished
way to implement an ALL-IN-1 application in TeamLinks. It also provides a nice
application which is useful and popular.
The DECUServe Journal September, 1995 Page 13
TeamLinks SDK
The CVRMEMO tool in the \CVRMEMO sub-directory allows you to customize mail
cover memos. This allows sites to establish their own unique mail message
formats just as is most popular in ALL-IN-1.
The final application of special note is called TL3BTN in the \TLEASY
sub-directory. The TL3BTN refers to the 3 active buttons this application uses
to control this application that bills itself as "TeamLinks Simple Mail". Not
only does it allow people to perform the most common mail functions in a very
direct and simple manner but it provides a view how one can perform a range of
key operations.
Making a submission to the SDK
The SDK
Submissions for inclusion in the SDK are always welcome. New submissions
should demonstrate how to use any of the TeamLinks APIs in a way that hasn't
been done before. Send the source code for your submissions on a floppy
diskette to:
(External to Digital) Tony Redmond
Digital Equipment Corporation
Park House
North Circular Road
Dublin 7, Ireland
Internet Tony.Redmond@dbo.mts.dec.com
Rumour is that a large European customer is submitting some nice applications
they have developed to help their clients migrate from ALL-IN-1 IOS to
TeamLinks. If you have a good example, sending it in can help others.
SDK Sub-directories
Below are the SDK sub-directories in alphabetic order as stolen
from the primary SDK documentation that escaped in June 1995:
AIDA Sample programs showing how to request and fetch data held on
an ALL-IN-1 server.
APPBMP DLL source code to show how application-specific icons can be
created and stored.
CONNECT Sample program to show how to use the TeamLinks connection
cache.
DIRECT Sample program to show how to use TeamLinks directory services.
DOC Documentation for APIs and other information relating to
TeamLinks.
EXTDLL DLL source code for many different types of functions that
extend TeamLinks functionality.
The DECUServe Journal September, 1995 Page 14
TeamLinks SDK
FCCABSPY DLL source code for a program used to monitor event processing
between TeamLinks and applications using FILECAB.DLL.
FILECAB Many sample programs illustrating how to use FILECAB.DLL and
CFC.DLL.
IAD Sample application showing how to list available drawers on an
ALL-IN-1 system.
IOSP Sample application showing how to use the script symbiont and
AIDA server to print documents using ALL-IN-1 resources.
MAIL Sample programs showing how to use the TeamLinks DDE
functionality.
MAPI DLL to emulate Simple MAPI calls.
MSAPPS Integrations for Microsoft applications offering new or
extended functionality over the standard TeamLinks components.
Applications include Access V2.0, Word V6.0, Excel V5.0, and
Project V3.0.
PCBROAD Utility programs that can be used to test broadcast
transmissions to PCs.
PCNOTIFY Sample application to manage the contents of the OA$PC_NOTIFY
data set on an ALL-IN-1 server.
PMEMOS Project showing how to use TeamLinks APIs to create a phone
memos application.
REFORMAT Source code for DLL functions that reformat text in the cover
note text box.
SHARED Shared global modules for use with Visual Basic programs.
TEAMLINK Software licenses and other information about the product.
TEAMROUT Sample code that can be used to build TeamRoute applications.
TLEASY Source code for "TeamLinks Easy", a 4-button implementation of
TeamLinks!
TLINET Source code for utilities that make TeamLinks a little more
"Internet aware".
TLSAMPLE Source for a C program that exercises many of the TeamLinks
file cabinet functions.
VIEWDL Source code for a DLL function to expand and display the
contents of a distribution list from the Create Message window
WPSCOPY Source code for the utility program used to import files from
WPS-PLUS for DOS file cabinets into TeamLinks.
The DECUServe Journal September, 1995 Page 15
TeamLinks SDK
Aliases for ALL-IN-1 Account Names
------- --- -------- ------- -----
Abstract:
A discussion of changing ALL-IN-1 account names, specifically the
problem of handling e-mail for the old address. One participant
suggests DDS, which leads to a follow-up on an interesting method of
maintaining the DDS database.
Participants: Bruce Bowler, John Briggs,
Gary Burke, Laurie Maytrott.
Conference: ALL-IN-1
Note 1145.0, 9-Aug-1995
Maytrott: Where can I set an alias for an All-in-1 account name?
----------------------------------------------------------------
We're going through and standardizing (read "shortening") our longer
usernames at my site. Is there a way we can have All-in-1 know that
mail sent to MENICHINO should now go to MENICHIN, for example, without
each user having to set nicknames to do this? Is there an option the
user whose name has been shortened could use to do this? My guess is
that the user-takes-care-of-it option is not available due to the
security risk of someone "hijacking" some else's mail.
Note 1145.1, 9-Aug-1995
Bowler: A&Q
-----------
You could create a profile record for MENICHINO that autoforwards to
MENICHIN. There are some problems with this, for example, if MENICHIN
autoforwards their mail, mail addresses to MENICHINO won't make it past
MENICHIN (ALL-IN-1 only autoforwards once to prevent loops).
Why are you "standardizing"? Is it because you have a system
"somewhere" that has an 8 character username limit? Are you aware
that ALL-IN-1 username and VMS username don't have to be the same?
The DECUServe Journal September, 1995 Page 16
Aliases for ALL-IN-1 Account Names
Note 1145.2, 9-Aug-1995
Briggs: DDS lookups can handle long usernames
---------------------------------------------
If you're using DDS for your directory service, then references to
the person's complete surname should get mapped properly -- user keys
in MENICHINO at the TO: prompt and ALL-IN-1 does a directory lookup,
discovering that MENICHINO points to MENICHIN.
We use eight character usernames here and use the DDS for directory lookups.
I just checked and verified that the behavior is as described above.
Note: We maintain the DDS directory ourselves and don't let ALL-IN-1
get its grubby little paws in there. The program that does the maintenance
makes sure that everyone has two surnames -- their username and their true
surname (if they are different).
Note 1145.3, 10-Aug-1995
Burke: How are you maintaining DDS?
-----------------------------------
John, I have an interest in knowing what other ways folks are
maintaining DDS databases (other than ALL-IN-1 and DSU). It sounds like
you may be doing something different?
Note 1145.4, 11-Aug-1995
Briggs: DDS update via Fortran and MBMAN
----------------------------------------
We maintain a separate database (using Datatrieve and an RMS indexed file).
Every night, I run a Fortran program that looks for changes in the
RMS database and creates an MBMAN script to update the DDS.
The program updates all the standard fields (names, initials, address,
phone, location, MHSORADDRESS/VMSORADDRESS, etc). It also updates one
of the user fields (for use by our mail hub). We do the stupid
trick of parsing a one-piece full name to produce given name, middle
initial, surname, generation and title of respect. That part works
surprisingly well. And, as mentioned in my previous reply, we set
both the user's last name and his e-mail ID as surnames. And we set
a unique UAID to distinguish our entries from those put in by any other
user agent.
I catch modifications to the RMS file by saving last night's copy
and doing a record-by-record compare with this night's copy before saving
the new file.
I do a sanity check for deletions by comparing the output of
MBMAN>SHOW DDS SUBSCRIBER /SUBID=* against the RMS file.
It's all pretty much brute-force. We don't have any MRIF tools.
It works well -- it's been six months since I've had to go tweak
anything.
The DECUServe Journal September, 1995 Page 17
Aliases for ALL-IN-1 Account Names
Note 1145.5, 11-Aug-1995
Burke: Good info!
-----------------
Thanks, John. That will be helpful. We're becoming the central routing
point for all ALL-IN-1 mail for all the other KP ALL-IN-1 using regions
(about 5) and also will handle dir-synch for them, so I'm looking for
simple and best solutions. The reason for all of this is a top-down
decision to use Mail*Hub for a wheel-and-spoke type interregional mail
system. The implementation is being pushed through under sever time
constraints which are having very negative effects on the way things
are setup. They wanted us to give them access to a privileged account
so their software could update our DDS directly. We said, as politely
as we could, "No, thank you very much. Just send us a flat file and
we'll send you the same". We're supposed to test the inbound (to them)
synch next week, and I'm just getting into DSU, so ... more fun. Keeps
life interesting, though.
PMDF and Message Router
---- --- ------- ------
Abstract:
A discussion of using PMDF's Message Router replacement for
VAX-to-Alpha migration of ALL-IN-1; one site's experience, and
lessons learned.
Participants: Michael Connelly, Mike Lampson,
Lynda Peach.
Conference: ALL-IN-1
Note 1152.0, 24-Aug-1995
Peach: PMDF V5.0, no MsgRtr, PMDF-LAN
-------------------------------------
Has anyone upgrade to PMDF 5.0? And if so, did you do so by selecting
the configuration which no longer requires Message Router?
Also PMDF-LAN? Is anyone running that?
WE have found a bug in our current version of PMDF Version 4.3-10 and
even upgrading the PMDFSHR.EXE patch did not resolve it. The only
solution is to upgrade. And if we do that we want to stop using the
VAX (we run ALL-IN-1 on alphas) and would like to take advantage of the
PMDF-LAN.
The DECUServe Journal September, 1995 Page 18
PMDF and Message Router
I'm open to any suggestions, guidance, comments.
Oh, the bug -- no mail moving out of the PMDF mailbox and the PMDF
mr_local process taking as much as 100% of the CPU resulting in a
CPU-bound system.
Note 1152.1, 27-Aug-1995
Mark Lampson: work in progress ...
----------------------------------
I was helping a client set up PMDF-MR sans MR last week. We got
it working after a couple of hours. But then when they went to
move the rest of ALL-IN-1 off the VAX this weekend, the Sender
(but not the Fetcher) stopped working. Still no resolution on
this as I just found out about it a couple of minutes ago.
This particular situation had an ALL-IN-1 system communicating
with the Internet. No other gateways or ALL-IN-1 systems to worry
about.
Note 1152.2, 28-Aug-1995
Peach: Keep me posted
---------------------
Hmmmmm......
Our Alphas are running ALL-IN-1 and the only thing the VAX runs is
Message Router. Do you think what you are/were dealing with would
happen to us?
BTW, right now mail is moving fine so maybe the patch *did* help.
Monday is always our worst mail day.
Note 1152.3, 28-Aug-1995
Lampson: still in progress ...
------------------------------
Maybe. Are PMDF's Batch queues currently on the VAX or Alphas?
My client's problem is still not resolved, but it seems that it
has something to do with moving the PMDF batch jobs from the VAX
to the Alpha.
As you may guess, it isn't necessarily straight-forward - at least
for VAX to Alpha moves. I would guess that if PMDF had originally
been installed on the Alpha and not upgraded on the VAX, things
would be easier. The documentation doesn't really explain any of
this. It just shows an example of configuring with the MR
replacement.
The DECUServe Journal September, 1995 Page 19
PMDF and Message Router
Note 1152.4, 28-Aug-1995
Peach: PMDF on VAX - wanna go to Alpha
--------------------------------------
PMDF queues are on the VAX but seen on the Alphas and I can do a
stop/que/rest from a priv'd account on the Alphas.
Have you checked on PMDF's www to see what they have to say? I have to
admit I haven't -- I'm still in the initial ck-it-out phase.
Note 1152.5, 29-Aug-1995
Lampson:
---------
No, I haven't checked out the WWW pages. I'm not directly
involved in fixing the current problems. Last report was
that PMDF Support was "stumped".
Note 1152.6, 31-Aug-1995
Connelly: PMDF-MR works!
------------------------
Well, we're the ones Mike Lampson has been working with and, yes,
PMDF-MR is now working beautifully as a replacement for Msg Rtr. Our
Alpha-based ALL-IN-1 v3.1 email is flowing smoothly to and from the
Internet.
Here's the scenario - we're migrating from VAX to Alpha (both
clustered) and Msg Router based email from ALL-IN-1 to PMDF-MR based.
Important Lesson: Don't try to be too smart. We "tried" to keep Msg Router
running on the VAX where our production users were while trying to
bring up PMDF-MR to replace Msg Rtr on the Alpha installation of
ALL-IN-1.
Where we thought we had it working, it seems that PMDF-MR was
configured just smart enough to take the email and hand it off to the
Msg Router on the VAX. Since the mail was moving, we gratefully
assumed that we had successfully installed PMDF-MR. Dumb! Later, when we
then turned off Msg Rtr, the email stopped flowing. Our problems were
further confused since we were also migrating batch and print queues as
well as the ALL-IN-1 directory tree from the VAX to the Alpha.
Unless you really are on top of this stuff, trust the folks
at Innosoft (PMDF) and keep things simple. Once we understood what was
going on, a simple change to the PMDF config file was all that was
needed. Innosoft was a great help in this effort (although we did have
them stumped when we swore to them that we had it working for awhile
but then it stopped).
Good Luck. I have absolutely no hesitation recommending the PMDF-MR
route. PMDF is an excellent product and the folks at Innosoft are
definitely on top of their game. (and my only interest is that of a
The DECUServe Journal September, 1995 Page 20
PMDF and Message Router
satisfied customer).
Note 1152.7, 31-Aug-1995
Peach: Vunderbar!
-----------------
That is great news. I have confidence that it would all work out.
I too have had wonderful experiences working with the Innosoft support
group. Thanks for your info!!!
Reverse Linking of Shareable Images
------- ------- -- --------- ------
Abstract:
If you need to link with two shareable images, and routines in each
of the shareable images depend on the other, you have what is
commonly known as "a problem." Although sometimes these situations
"just happen," there may also be good reasons for wanting to do
this. One approach to solving the problem is circular linking of
the shareables. Other possible solutions and workarounds are
described as well.
Participants:
Milton Campbell, Henry Carmichael, Chris Wesling, Ray Whitmer.
Conference: DEC_SOFTWARE
Note 700.0, 5-Jul-1995
Carmichael: Reverse linking of Shareable Images
-----------------------------------------------
"Double Reverse Linking" for Shareable Images
I have looked for any references to the title subject in
DECUServe, but have not found any. Perhaps someone can point me
in the right direction. I have need of the functionality on both
VAX (V5.5-2) and AXP (V6.1).
We use shareable images extensively in our VMS software. As
currently implemented, if an executable image is linked with shareable
images A and B, the main routine can call a routine in shareable A,
which can, in its turn, call a routine in shareable B. However, the
routine in shareable B cannot then call another routine in shareable A.
There is supposed to be a way to get around this difficulty,
The DECUServe Journal September, 1995 Page 21
Reverse Linking of Shareable Images
which we have called "double reverse linking". Briefly, it involves
1. Building shareable image A, which will have the routine in
shareable B as an undefined symbol.
2. Building shareable B and including shareable A, so the
routine in A is known to shareable B.
3. Rebuilding shareable A and now including shareable B, so the
previously undefined symbol is now resolved.
It sounds reasonable, but we have not been able to make it
work. Have any of the VMS gurus or novices created a working "double
reverse linkage, or has anyone heard of it being done?
Note 700.1, 5-Jul-1995
Whitmer: More questions.
------------------------
It _almost_ defeats any purpose of separate RTLs if they are statically
linked to each other. Object oriented analysis seems to assert that if
there is a one-to-one unconditional relationship between two
objects/entities, then you should strongly consider combining the
objects. But this is only the analysis phase, not the design phase,
which says to do it however it makes sense. I can think of reasons to
keep the RTLs separate -- it might make it easier to replace one or the
other either because of problems creating the thing as a whole all the
time.
I do not remember explicitly, but it seems like I may have tried
reverse static linking like you are trying before and failed -- it just
feels wrong to me. But if you make all your transfer vectors correctly,
I cannot think of a specific reason why it would not work. The linker
likes to keep track of images that had errors as they built, and I
think that the error originated by the undefined symbols would probably
be forever propagated between images as they link to each other unless
you patch success into one of the images after the undefined symbols
are gone. I think you need to tell us exactly what errors you are
getting which prevent this from working. Also, are both images
unbased, i.e. completely relocatable? Also, how many symbols are we
talking about that typically need to be resolved? What kind of
symbols? Only routines? etc.
In this type of situation I have always opted for dynamic activation,
like via LIB$FIND_IMAGE_SYMBOL. In cases of tight back-linking like
you are doing, I usually passed an array of back link pointers in an
initialization call from one DLL to the other, which then places it
into a global variable. Then I caused the RTL receiving the pointer
array to reference indirect (*) through the appropriate offset in the
array (the direction with the least dependencies). When there were
many interdependencies to manage manually, I somewhat automated the
process of binding the array on one side and using it on the other.
The DECUServe Journal September, 1995 Page 22
Reverse Linking of Shareable Images
Note 700.2, 5-Jul-1995
Wesling: Circular ("double reverse") linking not recommended
------------------------------------------------------------
We used to have some RTLs that were circularly linked to each other,
more by accident than by design. We had to redo them to eliminate the
circular calls when we began porting to AXP -- as I recall, the AXP
linker just refused to do it at all.
Generally, if image A could call image B's routines, we either moved
routines from A to B so B wouldn't have to call A any more, or made
image B use LIB$FIND_IMAGE_SYMBOL to call image A's routines (which
eliminates the need to link B to A).
Eliminating circular linking also makes it much easier to install or
upgrade the RTLs... we used to have to link image A as far as it would
go, then link image B to what there was of image A, then relink image A
to image B, or something like that. With one-way dependencies, all you
have to do is link the first image, then link the image that depends on
the first one -- *much* simpler process all around.
Note 700.3, 5-Jul-1995
Campbell: Maybe this would work
-------------------------------
While I have to join in not thinking this is a good idea. It seems
like you might be able to short circuit some of the error stuff by
linking image B first with dummy entry point for the A routines it calls.
This should make the linker happy with B. Then link A (against B). A
now is "real". Remove the dummy entry points from B, relink it against
A. It seems like this might work, but the end result still looks like
a wonderful booby trap for a future maintainer.
Note 700.4, 7-Jul-1995
Carmichael: ...maybe it would. I'll try it.
--------------------------------------------
We have separate shareables due to the wish to keep "universal"
routines (shareable A) separated from "database specific" routines (shareable
B). Each "flavor" of B will have the same routine names, but variable content
depending on the database to be accessed (RMS, Vendor X relational database),
as determined by the needs of the user. Applications will use the universal
routines (A) for their generic needs, and access the database through generic
routines. Database specific routines (B) will do what is needed on a lower
level and return the result to the application. If a routine in B needs
generic information, it can (we hope), call an A routine to get what it needs,
thus, the reverse linking.
WRT Ray's questions (.1), yes, both sharables are completely relocatable.
We are talking about perhaps five symbols, routines only, which need
resolution.
I think that Milt's approach (.3) looks promising. I plan to try it,
The DECUServe Journal September, 1995 Page 23
Reverse Linking of Shareable Images
and get back with results. I hope that the booby traps will all have been
sprung (wishful thinking?) before a future maintainer comes along. As long as
I am the booby, it's not too bad. My first try will be on AXP, since that's a
more immediate need. VAX will follow, even if AXP doesn't seem to work.
There may be a better way to accomplish this with an object oriented
approach, but I don't know enough to approach it from that direction. If
anyone could point to the path of OO righteousness, I would appreciate it.
Note 700.5, 10-Jul-1995
Wesling: How about routine callbacks?
-------------------------------------
Have you considered passing routine addresses to use as callbacks? The
CMS callable interface uses that method very successfully to avoid
exactly this kind of problem.
Here's what it would look like in your situation:
* Application calls generic routine in library A.
* Generic routine calls database specific routine in library B.
* One of the parameters that the A routine passes to the B routine is
the address of another A routine that can be called by the B routine
if/when it needs more generic info.
This allows the B routine to call the A routine without having to link
the B library to the A library.
Having only used this method in situations where the B library was a
DEC-supplied library, I'm not sure what gyrations your B routine would
have to go through to call a routine given its address. In MACRO, I
think it would just be a JSB to that address; in high-level languages,
it would depend on the language. C should be able to handle it pretty
easily, I believe.
DCPS Memory Leak Fixed
---- ------ ---- -----
Abstract:
In the past, a memory leak "feature" was discussed in the Notes
topic dedicated to DECprint Supervisor. Recently, a subscriber
asked for an update. In case you were wondering, too, here's the
answer: it's fixed.
The DECUServe Journal September, 1995 Page 24
DCPS Memory Leak Fixed
Participants:
Bob Graham, John Osudar, Alberto Zuniga.
Conference: DEC_SOFTWARE
Note 580.109, 23-Aug-1995
Zuniga: Version 1.2 ...bug fixes?
---------------------------------
Does anyone know if the memory munching bug in DCPS V1.1 has been
corrected yet. I noticed that V1.2 is out, but I don't have access to
the release notes.
Note 580.110, 24-Aug-1995
Graham:
--------
Yes, the various memory "leaks" in v1.1 and v1.1A are supposed to
be fixed in v1.2. I've been running v1.2 for over a month now with
50+ DCPS execution queues and haven't seen any noticeable memory
creep in the symbionts.
In addition, the v1.2 symbiont can be run multistreamed (more than one
execution queue per symbiont process). By default, you get only the
single streamed behavior, but you can define the logical
DCPS$MAX_STREAMS to set the number of queues per processes. According
to the release notes, setting the max streams to at least 4 should
result in about the same memory consumption as DCPS v1.0C (fewer, but
larger symbiont processes).
Note 580.111, 24-Aug-1995
Osudar: multistream 1.2 works fine
----------------------------------
We're running DCPS 1.2 with 4 streams per process, with no problems.
btw Digital says that 1.1 didn't have memory leaks (at least not
unintentional ones). And multistream 1.2 symbionts still have to
serialize translator access -- so if you use translators and one of the
queues run by a multistream symbiont is translating a huge file, the
others can't translate anything until it's done (but they can print
untranslated jobs just fine).
800 bpi Tape Drive Needed
--- --- ---- ----- ------
The DECUServe Journal September, 1995 Page 25
800 bpi Tape Drive Needed
Abstract:
Some older equipment can be challenging to replace. The querent in
the following topic has a need to read tapes at 800 bpi, which is
lower than most currently-existing drives will handle. Not a
problem for DECUServe, though.
Participants: Barton Bruce,
Lesley Gray, Terry Kennedy, Pete Sivia.
Conference: HARDWARE_HELP
Note 1907.0, 19-Jul-1995
Gray: Do you have an 800bpi tape drive?
---------------------------------------
Help...
I am in desperate need of an 800bpi tape drive. I'm willing to borrow, buy,
rent, lease, etc. within reason. We only need the 800 bpi tape drive for about
5 months. Then the whole computer complex is being upgraded. We have an older
flight simulator that compiles it's code off of an 800bpi tape drive and the
two that we have on our VAX have both died. We've spent thousands trying to
repair them with no success. We've tried to lease new ones but so far we've had
one manufacturer drop out after we placed an order 2 months ago and another one
that won't guarantee it will work for twice as much money.
Any help or suggestions would be greatly appreciated!
Note 1907.1, 19-Jul-1995
Sivia: Did you try the used equipment dealers?
----------------------------------------------
Umm, you don't say **what** kind of drive interface you need, but try
somebody like Midwest Systems (1-612-894-4020) or Newman
(1-313-426-3200) see what they have used. Newman used to (about 2
years ago) have a huge warehouse full of older gear but what you want
predates the old Tx78/79/81 families by a whole density so it might be
tough...
(No connection, just know that folks have bought good stuff from both
of them in the past)
Note 1907.2, 19-Jul-1995
Kennedy:
---------
I don't think you want a TU77 (or earlier generation) drive. They're much
too big and flakey. I have a Kennedy something-or-other (no relation) which
The DECUServe Journal September, 1995 Page 26
800 bpi Tape Drive Needed
claims to do 800/1600/3200/6250 and be SCSI. I bought it used and I have no
manuals. However, doing 800 is trivial on a modern "pizza oven" drive - it's
about 1K of extra software (only - no hardware) in the drive. I'd suggest
calling Overland Data or Qualstar - look in the back of PC Magazine in the
red classifieds under "media conversion" or "9 track".
Note 1907.3, 21-Jul-1995
Gray: Thanks!
-------------
Thanks for all the hints!!!
I'll let you know if I find one. If not, I'll be replacing the old computer
system way ahead of schedule. :-)
Note 1907.4, 22-Jul-1995
Bruce:
-------
> too big and flakey. I have a Kennedy something-or-other (no relation) which
> claims to do 800/1600/3200/6250 and be SCSI. I bought it used and I have no
Probably 9612, or more common 9610 without SCSI.
Should be very cheap on the used market. If you have a q-bus vax, you can get
used tape controller so you won't need the scsi version. There
are many notes on decuserve that talk about these. If you can
make forwords behave, look for DQ152, DQ153, DQ142, DQ143 or DQ132. Some of
those don't exist, but I can't remember which. Those are dilog numbers.
there are good emulex ones, too.
Note 1907.5, 9-Aug-1995
Gray: Got one. Thanks!!
-----------------------
Thanks for all the help. I ended up with a new Kennedy 9612. It was a bit
pricey but I'll let you know if it was worth it after it arrives. :-)
Anything to get the Advanced Maneuvers up and flying in the sims!
Video for Alpha 2000-300
----- --- ----- --------
The DECUServe Journal September, 1995 Page 27
Video for Alpha 2000-300
Abstract:
Exactly what is "cheap" varies considerably. Putting together a
system for home use tends to impose rather tight budget constraints.
If you have your eye on an Alpha, a used 2000-300 (previously known
as the AXP PC 150) might start to look pretty good. Video options,
as discussed in the following notes, are limited, though.
Participants:
Duncan Brown, James Laferriere, Pete Sivia.
Conference: HARDWARE_HELP
Note 1915.0, 14-Aug-1995
Brown: Video options for the 2000-300 (AXP 150)
-----------------------------------------------
The AXP 2000-300 isn't even sold anymore, but I'm looking at buying an
AXP machine running VMS as a "home PC", and am trying to weigh all my
options.
All of the S&O Catalog type materials I have on the 2300 only ever
mention the Qvision card, which is supposed to only be 1024x768. I
want to run a 1280x1024 setup with a 19" monitor. Is there a card
available (and supported under VMS) that will allow me to do this? Or
do I have to scratch the 2300 from my list of potential machines?
Note 1915.1, 15-Aug-1995
Sivia: Qvision is the one I've used
-----------------------------------
The Qvision card is the only one I ever used on the -300 and it's the
only one that works with VMS according to the January 1995 S&O. The
cards available:
PB2GA-AA 8bit EISA Qvision. 1024x768 (VMS, NT)
PB2GA-CA 8bit ISA Number 9 GXE 1024x768, 1280x1024 (NT)
PB2GA-EA 24plane ISA Matrox MGAI 1280x1024 (NT)
I've got some notes back in my office from a Spring '95 Symposium talk
by George ? who was/is doing DEC's video card driver work in this area.
I'll find them later today and post what he had to said about
additional graphics card support.
FYI: if you are interested in a used DECpc 150 with NT license,
Qvision, 32MB, E'net, 500MB or 1GB disk, send email. I think I know of
one in excellent shape that the owner would sell at a quite reasonable
price. This system has also run VMS using the SDK kits.
The DECUServe Journal September, 1995 Page 28
Video for Alpha 2000-300
Note 1915.2, 15-Aug-1995
Brown: Close, but no PCIgar...
------------------------------
> The Qvision card is the only one I ever used on the -300 and it's the
> only one that works with VMS according to the January 1995 S&O. The
> cards available:
..which is what I was afraid of...but I may have just stumbled on
something with my favorite friend, the wildcarded part number search on
the E-store:
-PB2GA-FA mach64 2D gfx, all OS's E/U Adj. %
Service: 1 Year 2 Years 3Years
Gross Net Gross Net Gross Net
List: 399.00
Standard:
Description:
ATI mach64 WinTurbo "CX" (#30-44635-01), 2MB DRAM PCI
graphics accelerator, to 1280x1024x8, all OS's.
..OH DRAT, I just remembered: no PCI bus in the AXP 150. Oh well. I
figured since the part number format was similar it would work.
Sigh.
I wish DEC would come out with a PCI version of that computer for under
$1000...yeah, that's the ticket!
Note 1915.3, 16-Aug-1995
Sivia: Notes from Spring '95 Symposium
--------------------------------------
I found my notes from the Spring '95 Symposium presentation on AXP
graphics support given by Bill Hanley (OpenVMS Product Management) who
was subbing for Fred Kleinsorge. Fred's doing the actual graphics
adapter work and periodically appears in the news groups. Maybe you
might consider sending him email at FRED@EXTASY.ENET.DEC.COM to see if
he's got any additional info/ideas for you.
Anyway, a couple of the notes I took during this session:
- VMS V6.2 supports up to 8 of the Qvision cards for multi-head
support. Using the EISA Qvision variant, this is one way to get more
real estate for additional windows if that's what you wanted to do with
the larger 19" monitor at 1280x1024 resolution. (Obviously if you
wanted 1280x1024 in order to see the detail in one window, this option
doesn't help.)
- VMS V6.2 also supports multihead ATI Mach 64 CX PCI cards (2MB DRAM
variants only). Does require new console firmware (which I'd guess is
on the V3.2 firmware CD).
- VMS V6.2 also supports the ATI Match 64 GX cards which the presenter
The DECUServe Journal September, 1995 Page 29
Video for Alpha 2000-300
said were better performers, but were in limited supply which is why
DEC didn't sell them.
- During "Summer '95" S3 Vision 864, Trio32, Trio64, DEC864 chip cards
were to be supported in VMS (needs a patch kit).
- Maybe late CY95 there would be support for large virtual screens such
as 1280x1024 using a 640x480 monitor.
Note 1915.4, 16-Aug-1995
Brown: Has some promise
-----------------------
> - VMS V6.2 supports up to 8 of the Qvision cards for multi-head
> support. Using the EISA Qvision variant, this is one way to get more
This is actually not a bad fix. I do mostly just want the real estate
for more windows, not bigger ones. Two problems, but I can probably
live with them:
-- More "real" real estate consumed on my desk
-- "Flight" assumes 1280x1024 (heh heh heh...)
> - During "Summer '95" S3 Vision 864, Trio32, Trio64, DEC864 chip cards
> were to be supported in VMS (needs a patch kit).
Are any of those EISA-based?
I'll probably drop Fred a line and see what else he has to say.
Note 1915.5, 16-Aug-1995
Laferriere: Compaq used 'em on system-pro series.
-------------------------------------------------
Duncan, I am quite certain that there are eisa variants of the
Qvision cards as Compaq used them on some/most of their eisa
pc's ...
Note 1915.6, 17-Aug-1995
Sivia: DEC's Qvision is EISA
----------------------------
Yes, DEC's Qvision card is EISA. See note .1 for the part number.
Note 1915.7, 17-Aug-1995
Brown: From the Fred's mouth
----------------------------
Fred Kleinsorge sent me a nice return e-mail and explained why we're
never going to see 1280x1024 support under VMS for that machine.
The DECUServe Journal September, 1995 Page 30
Video for Alpha 2000-300
Amongst other reasons:
-- There's a 1280x1024 Qvision card but it's at end-of-life so they
don't feel the need to support it. (Let's never mind that the 2000/300
itself is no longer sold!)
-- Limited memory support under VMS in the 2000/300 architecture
(bigger screen = more contiguous memory required, of course!)
He suggested a much better choice of machine would be one of the new
small PCI ones, eg the Alphastation 4/100 or 4/266, etc. No kidding!
DEC is really making some amazingly small, powerful, cheap machines
these days. Since my interest is for home use I need to get it even
cheaper than cheap, though, and only the 2000/300 seems to be getting
down into the right price range...
(Maybe I should convince the group here with a 2000/300 to dump that
old obsolete hardware and replace it with a newer Alphastation, heh heh
heh....)
SO the definitive answer would seem to be:
1024x768 is high as you're going to get on a 2000/300. With VMS 6.2
you can add multiple 1024x768 screens but that's it.
RAID Options on OpenVMS
---- ------- -- -------
Abstract:
Which flavor of RAID makes the most sense will depend on the system
and requirements in question. An analysis of one situation,
compliments of one of Keith Parris in the StorageWorks group,
follows.
Participants:
Keith Parris, Kent Poots.
Conference: HARDWARE_HELP
Note 1921.0, 21-Aug-1995
Poots: VMS Shadowing Options ??
-------------------------------
Any thoughts on the best way to RAID some Oracle database disks ?
Choices are (under VMS) HSJ-based controller, or host-based
(host is VAX currently, will be Alpha shortly).
Feedback appreciated.
The DECUServe Journal September, 1995 Page 31
RAID Options on OpenVMS
Note 1921.1, 22-Aug-1995
Parris:
--------
You'll need to decide first what your goals are in using RAID:
o Better performance? (RAID-0)
o Better availability? (RAID-1 or RAID-5)
- RAID-5 costs less; RAID-1 has better write performance
o Both? (RAID 0+1)
In making this decision, it helps to know:
o the read/write ratio
o the average I/O size
Do you already own any of the pieces, i.e. an HSJ controller, a set of disk
drives, the host-based RAID software or host-based Volume Shadowing?
Note 1921.2, 23-Aug-1995
Poots: Thanks ! - HSJ Questions...
----------------------------------
Thanks for your reply.
Got an HSJ40, plus a spare on the shelf. Goals are performance and
availability. Can buy more disks (plan to ).
This is primarily an Oracle - based set of applications.
Right now we've got 5 busy database disks. 3 on one SCSI channel and
2 on another SCSI channel. Not sure if HSJ40 is intended to overlap
reads and writes on 1 channel very well. (I inherited this config !!).
VPA complains of queues, so something is up.
I'm also not quite sure how disks are normally allocated for RAID
on an HSJ - spread members across different channels, I assume ?
Thanks in advance for further comment !!
Note 1921.3, 24-Aug-1995
Parris:
--------
With both performance _and_ availability important, one would normally go for
RAID 0+1 instead of RAID 5. However, if you have write-back cache in the
HSJ40, and you aren't running near saturation, we're finding the write-back
cache hides the extra RAID-5 write latency so well that the write performance
is almost indistinguishable from the host's viewpoint.
> This is primarily an Oracle - based set of applications.
With Oracle, writes will probably tend to be around 16 blocks average size, a
bit higher than the typical 4-8 block average I/O size for VMS in general, but
The DECUServe Journal September, 1995 Page 32
RAID Options on OpenVMS
it's not so large as to be able to use multiple disks at once in a single
transfer, so the performance advantage in using RAID is that you'll be
spreading the I/Os evenly across the available spindles (no "hot" spindles, and
as many spindles as you need to handle the load) for lower average response
times. You'll just want to choose chunk sizes that are a bit larger than what
you'd choose for typical VMS I/O sizes (say at least 16*20=320 blocks vs.
4*20=80 blocks).
> Right now we've got 5 busy database disks. 3 on one SCSI channel and
> 2 on another SCSI channel. Not sure if HSJ40 is intended to overlap
> reads and writes on 1 channel very well. (I inherited this config !!).
The HSJ40 supports disconnect/reconnect on the SCSI busses, so it can overlap
seeks and data transfers among multiple disks on each SCSI bus. With typical
Oracle I/O sizes, you're most likely spending much more of your time seeking
than transferring.
> VPA complains of queues, so something is up.
Could just be that the individual disks are plugging away as fast as they
can, and adding more spindles is the answer.
> I'm also not quite sure how disks are normally allocated for RAID
> on an HSJ - spread members across different channels, I assume ?
As a rule of thumb, one would normally spread them across different SCSI busses
on a round-robin basis. You have 6 busses in an HSJ40 -- might as well use
them all, since you've already paid for them. :-)
Note 1921.4, 24-Aug-1995
Poots: 2 HSJ's; Other Tuning Hints ?
------------------------------------
Thanks for your feedback !!
RAID 0+1 using 2 HSJ40's will present any complications ?
Any specific testing/performance measurements which might be
appropriate before I go ahead and order several more disk drives ?
I'm pretty sure that there are application issues here BUT
I'm not a "big" Oracle person, and our applications group doesn't
have the resources to do radical things with the database and/or code.
(isn't this the lament of all system managers ?)
Note 1921.5, 25-Aug-1995
Parris:
--------
> RAID 0+1 using 2 HSJ40's will present any complications ?
None that I can think of -- putting two HSJs in a dual-redundant configuration
The DECUServe Journal September, 1995 Page 33
RAID Options on OpenVMS
just keeps an HSJ from being a single point of failure. Note that only one HSJ
at a time will service requests to a given RAIDset; but considering that we
don't think the HSJ is the bottleneck anyway, that shouldn't be of concern.
> Any specific testing/performance measurements which might be
> appropriate before I go ahead and order several more disk drives ?
Might be useful to see what the I/O rate and average I/O size are before you
conclude the disks are the bottleneck. The VTDPY utility on the HSJs might be
helpful here, or DECps on the host. (Another trick is to temporarily
reconfigure the cluster so the disks get MSCP-served and then use MONITOR MSCP
and SHOW DEVICE/SERVED to get some visibility of the I/O workload.)
UAS/UAF Merge
------- -----
Abstract:
A discussion of a PATHWORKS for OpenVMS ECO that allows merging of
an OpenVMS SYSUAF and a PATHWORKS UAS for the benefit of those folks
who are dealing with NT servers and trying to keep lots of user
accounts in synch.
Participants:
Dan Singleton, Michael Spatz.
Conference: PATHWORKS
Note 843.0, 18-Jul-1995
Spatz: UAS/UAF Merge
--------------------
Pathworks for OpenVMS ECO#2 and a patch which will allow a UAS/UAF merge
in an NT environment will be at our site on Monday.
Note 843.1, 19-Jul-1995
Singleton: More Details Please on UAS/UAF -> NT SAM
---------------------------------------------------
Mike, Are use saying that you can merge a UAS/UAF with a sam on an NT
server? I am getting ready to transfer all accounts on a VAX cluster
running PW Server 5.0 to an NT domain. I'm not sure what or how the
best way to do this is? I have about 300 users on the cluster. Any
additional information on this would be appreciated. TIA
The DECUServe Journal September, 1995 Page 34
UAS/UAF Merge
Note 843.2, 20-Jul-1995
Spatz: UAS/UAF Merge ??
-----------------------
I will pull together the documentation once it is received from Digital.
I can assume you are bringing your vax out of backward compatibility mode
and become a BDC in a Domain. If this is the case then the patch and ECO
Will make your life much easier. The MERGE as it stands now will corrupt
the UAF database in about 15 min after the merge. The ECO kit will set the
UAS and UAF passwords to be the same. Then a net password will give the user
the ability of keeping the UAS/UAF databases in sync. It is a nightmare
if the databases are out of sync. We pressured for this ECO since we have
in excess of 1000 accounts.
Note 843.3, 27-Jul-1995
Singleton: Interested in UAS -> SAM Merge
-----------------------------------------
I am looking for a merge of the Pathworks UAS and SAM or a merge of
the UAF and SAM. I'm not sure if this is the scenario that you are
describing in -.1? If so, I am definitely interested. TIA Dan
Note 843.4, 31-Jul-1995
Spatz: merge to NT
------------------
I expect to have the patch to ECO#2 later this week which will handle the
merge. This merge is needed due to the fact that the PDC is an NT server
which will corrupt the VMS passwords.
Note 843.5, 4-Aug-1995
Spatz: ECO#2
------------
Pathworks V5.01-ECO#2 is installed and in test mode. We are still waiting
for the patch to this ECO which will allow migration into an NT domain
without corrupting passwords. Currently they are having technical
problems generating the patch kit for ECO#2. Will keep you informed
during the NT merge in the near future.
Note 843.6, 10-Aug-1995
Spatz: ECO#2 update #1
----------------------
Just installed update #1 to Pathworks V5.0C ECO#2 which will allow
a migration into an existing domain with an NT as a PDC. The Patch will
allow the password to be set to the username. Migration into the domain
will be tested this afternoon. Stay tuned for the results.
The DECUServe Journal September, 1995 Page 35
UAS/UAF Merge
Note 843.7, 11-Aug-1995
Spatz: UAS Merge in an NT Domain
--------------------------------
Completed the UAF UAS merge with success. The ECO will set the UAS and
UAF passwords to be the same as the username that is entered into the NT
Domains database.
Then the user has the ability to change the passwords from LANMAN or VMS.
DHU-11 Installation Problem
------ ------------ -------
Abstract:
DECUServe is a good place for cutting-edge information, but it's
also a marvelous source of insight on systems that are, shall we
say, not quite so new any more. Here, there is a problem with a
newly-added DHU-11 module in an 11/84. And just what are those
other two connectors for, anyway?
Participants: Rob Brown, Arnold De Larisch,
Terry Kennedy.
Conference: PDP-11
Note 424.0, 9-Aug-1995
Brown: DHU-11 installation problem
----------------------------------
One of my customers is attempting to install a DHU-11 on a PDP-11/84
running RSX-11M-Plus V4.1.
He did parts of a sysgen to add support for the device. After a couple
of tries, he got the board installed so that the board's CSR matched
the one specified in the sysgen.
When he tried to test the DHU by broadcasting to one of its ports, all
of secondary pool was consumed. I had him disconnect all ports on the
distribution panel, try different ports, and even swap out the DHU
board. The problem persisted.
He is now doing a _complete_ sysgen to see if that straightens things
out. If that doesn't work, ... ?
Has anybody seen this failure mode before? I think we have eliminated
hardware as the problem here. Could I be mistaken? Is there a switch
on the board that makes it use up secondary pool?
The DECUServe Journal September, 1995 Page 36
DHU-11 Installation Problem
Note 424.1, 9-Aug-1995
De Larisch: Sounds like classic WHAT?-Huh? RSX Loopback error
-------------------------------------------------------------
This is a 'wild' stab in the dark ... but the broadcast to a port and all of
pool going away is a classic "loop back" problem.
The broadcast message is sent and MCR attempts to use it for login purposes ...
MCR sends out an error message which is then interpreted as a login attempt
(repeat until all of pool is exhausted with tt buffers).
I don't have any DHU's or the manuals for them ... but double check that they
aren't in "factory test mode" or something like that.
By the way, are you able to get CON to put the ports ONLINE? If you have a
two VT terminals ... you may want to hook them up to two of the ports and run
kermit to see if you can successfully transmit or receive info thru them.
Note 424.2, 9-Aug-1995
Brown: ... working ...
----------------------
> This is a 'wild' stab in the dark ... but the broadcast to a port and all of
> pool going away is a classic "loop back" problem.
Yes, I know. But where is it being looped back? That's why I asked
him to remove all of the cables from the DHU ports (whether or not he
did is another question, of course).
> I don't have any DHU's or the manuals for them ... but double check that they
> aren't in "factory test mode" or something like that.
That's a good idea. I don't have any manuals myself, but my customer
does. I'll have him check it out.
> By the way, are you able to get CON to put the ports ONLINE? If you have a
> two VT terminals ... you may want to hook them up to two of the ports and run
> kermit to see if you can successfully transmit or receive info thru them.
Yes, we can get them online. I had thought that broadcast was a good
enough test, but I see your point. Slaving a port so that MCR doesn't
get in the way could teach us something. Boy, I sure am out of
practice at this.
Note 424.3, 10-Aug-1995
Kennedy: Vector?
----------------
I don't speak RSX, but are you *sure* he got the csr/vector set right? If
the CSR is right and the vector is wrong, all sorts of funny stuff happens.
The DECUServe Journal September, 1995 Page 37
DHU-11 Installation Problem
Note 424.4, 10-Aug-1995
Brown: ... added to list ...
----------------------------
I'll have him double check. I can't think of a way to confirm the
vector on a running system, so I am not sure.
Note 424.5, 14-Aug-1995
Brown: Problem Solved!
----------------------
This problem turned out to be a hardware installation problem.
You are supposed to connect two ribbon cables to each H3029 distribution
panels -- one for each half-panel of four ports. There are two
connectors for ribbon cables coming from below, as well as two
connectors for ribbon cables coming from above.
My customer was using the upper pair of connectors. Switching the
cables to the lower set of connectors solved the problem.
If the distribution panels are supposed to work this way, I can't
imagine what the upper set of connectors is for.
Note 424.6, 15-Aug-1995
Kennedy: Loopback?
------------------
Newer DEC multiplexors (and I guess the DHU qualifies) put the loopback
connectors on the distribution panel(s). [Older ones shipped a separate
H-series loopback adapter which was always getting lost].
Note 424.7, 15-Aug-1995
Brown: Bingo!
-------------
The original problem was that everything seemed to be getting looped
back, so that would certainly explain it!
UNIX and DOS on the Same Machine
---- --- --- -- --- ---- -------
Abstract:
Various strategies for a multiple-OS PC; how to determine which one
gets loaded at boot time, access across file systems, and so on.
Featured buzzwords include Linux, OS/2, and Digital UNIX.
The DECUServe Journal September, 1995 Page 38
UNIX and DOS on the Same Machine
Participants: Petri Backstrom,
Andrew Hill, Matt Holdrege, Bart Lederman, Bradford Lubell, Benjamin
Pena, Kevin Roels.
Conference: PERSONAL_COMPUTING
Note 731.0, 9-Aug-1995
Holdrege: Unix and DOS coexistent
---------------------------------
I just read an article on Linux. It seems that I could partition my PC
disk and place a bootable Linux next to DOS. Has anyone tried this with
Linux or FreeBSD?
Another option would be to have two disks, Unix on one and DOS on the
other. How would I tell which one to boot?
Note 731.1, 9-Aug-1995
Hill: It can be done
--------------------
I've got two disks in my PC, and the second one is partitioned between
DOS and Linux (the first is just DOS). Booting is done via a program
called LILO (LInux LOader, I believe).
It's been a while since I configured it, but basically it
replaces the boot sector on your boot drive, and you can set it up to
boot DOS or Linux (or probably any other bootable system you've got).
There is a config file for setting up the boot menu. Somewhere it
keeps a copy of your original boot sector, but I don't remember where
that somewhere is.
Anyway, it's doable with either a single disk with multiple partitions,
or multiple disks. LILO is the key. BTW you can access your DOS
partitions from Linux.
Note 731.2, 9-Aug-1995
Roels: Can use the OS/2 boot manager also
-----------------------------------------
Some people here are doing Linux with OS/2, and DOS, in some cases.
They say they prefer the OS/2 boot manager to LILO.
Note 731.3, 9-Aug-1995
Lederman: Anyone tried the DOS environment within Linux?
--------------------------------------------------------
I've also read in the Linux blurbs (but haven't tried it) that there is
also a DOS environment or shell so you can run DOS programs under
Linux. It would be interesting to know if anyone has actually tried
The DECUServe Journal September, 1995 Page 39
UNIX and DOS on the Same Machine
it, though.
Note 731.4, 10-Aug-1995
Pena:
------
re: -2
Can use OS/2 Boot manager to get to LILO to whatever OS... (as sort of
a second chance to boot into other OSes than Linux at the LILO prompt)
or go directly into Linux.
Note 731.5, 10-Aug-1995
Pena: Linux isn't all that bad as a personal-use OS..
-----------------------------------------------------
re: -2
Program you're referring to is Dosemu... it's pretty nice, although on
a 486DX2/66 it's kind of on the....slow side. Although I haven't fully
fine-tuned it, though. But, worst case is a slow software-based
emulator.
Still... it does pretty good for what it does. I can pick up a floppy
with, say, Procomm, and connect to the VAX:
Linux->Dosemu->Procomm->VAX if I so desired.
As for reading OS/2 filesystems, under Linux, it's the 'hpfs'
filesystem, and to the best of my knowledge, it was read-only support.
Note 731.6, 11-Aug-1995
Lubell: Multiboot Windows/NT & OSF/1
------------------------------------
I have recently acquired an AlphaStation 250. It was pre-installed with
Digital Unix (formerly OSF/1). This computer has two (2) disk drives. The
users of this system would like to be able to boot either Unix or
Windows/NT. They do not want to dedicate one disk to each operating
system. What we would like to do is either share one of the drives between
both operating systems. Unix would be on both drives, whereas, Windows/NT
would only be on one drive. Does anybody know if this can be done? Does
anybody have any suggestions on where to ask?
Note 731.7, 12-Aug-1995
Backstrom:
-----------
You have to install different operating systems on
different disks on an Alpha.
The DECUServe Journal September, 1995 Page 40
POLYCENTER Console Manager
POLYCENTER Console Manager
---------- ------- -------
Abstract:
A discussion of the virtues of POLYCENTER Console Manager (PCM),
including several glowing customer testimonials. Also covered are
OpenVMS versus UNIX, RS-232 versus network, MTI Stingrays, and
filters available on the Freeware CD.
Participants: Rob Aldridge,
Larry Clegg, James Cristofero, Gary Gladstone, Frank Nagy, Pat
Scopelliti, Glenn Zorn.
Conference: VMS
Note 2561.0, 25-Jul-1995
Clegg: PolyCenter Console Manager
---------------------------------
Greetings Earthlings!
I'm attempting to bring in the Polycenter Console Manager hardware/software to
help reduce floor space in the datacenter for my five VAXen. I see that in
note #1784 there was a discussion of the old VCS system but the last update
was 25-Oct-92. I'm interested in anyone's experiences with this product as it
stands today under VMS V6.x . TIA.
Note 2561.1, 25-Jul-1995
Cristofero: USE VCS Extensively
-------------------------------
Not sure what you're looking for Larry. I have a VMS 6.1 system running
VCS 1.4 (prior to PCM). I have not upgraded yet as I'm awaiting memory
or a new box to run it on (PCM requires more memory of course).
VAX 3400 24 MB
1 RF73
2 RF71's one for VCS Logging one for user disk
Maintains 32 clients (Vax's, HSC's and Stringray controllers)
4 clusters in two facilities
4 4100 in four states
I have console access to a Vax in Indiana while sitting at home.
I've used this product since 1986. Have little complaints.
The DECUServe Journal September, 1995 Page 41
POLYCENTER Console Manager
Note 2561.2, 25-Jul-1995
Clegg: More Specifics
---------------------
I guess I should learn to be a little more specific in these questions. I'm
looking for things like: how has it helped/hurt in your system management
responsibilities? Now that you have it and knowing what you know about it
would you purchase this product again? What limitations, if any, have you run
into? Is there anything about it which surprised you - either positively or
negatively. I'm really looking for any "good ammunition" to help justify me
getting it - I *think* I want it but I want to be as sure as possible before
recommending it to my boss. I want to make a very informed decision since I
can't justify this purchase on any business application grounds - and just
saying I want it cause it will help me to better manage the systems won't cut
the mustard - that's way to nebulous. Thanks again.
Note 2561.3, 25-Jul-1995
Clegg: Yet More Specifics
-------------------------
I seem to really like to exit my editing sessions before I'm really ready!
More info:
My environment currently includes: 5 VAXen and 1 StingRay II, Infoserver 1000,
and soon an RS/6000 AIX system. This fall will see a new UPS for the
datacenter. There's also lots (250) PC's and some Novell servers. If I'm
reading the PCM SPD correctly - most, if not all, of this could be accessed
through the PCM.
Note 2561.4, 25-Jul-1995
Gladstone: It has saved Me many Miles
-------------------------------------
Real simple with the VAX's and PCM. We reboot our VAXs alternately on
Sundays (our only downtime window). This past Sunday, our mail cluster
was scheduled to be rebooted. Operations started the reboot on one of
the VAX's and it went into a fatal bug check loop. After receiving a
panic phone call at home (1 hours commute each way), I logged in and
connected to our VCS node and did a review of the vcs console log for
that system so I could see exactly what happened. I was also able to
connect to the console so I could do a conversational boot with verify
to see what the problem was.
It has paid off many times for this type of work and has saved many
miles commuting when there is a problem. We have one node that isn't on
VCS and I cringe every time there is a problem as I have no console logs
to review (kind of hard to read an LA75 printout with a bad ribbon from
60 miles away).
I could probably go on and on, but hopefully this gives you some info.
The DECUServe Journal September, 1995 Page 42
POLYCENTER Console Manager
Note 2561.5, 25-Jul-1995
Scopelliti: Don't forget the pro-active event notification
----------------------------------------------------------
Another big plus is the event notification system. Once you've been
saved by getting an alarm for "pagefile space critical" and having
the time to install another pagefile before the production system
hangs, you'll never want to be without it again.
Pat - who's been saved more than once from having to get dressed, drive
1/2 hour, hit "B <RETURN>", driving 1/2 hour, and getting back into bed
at 3:00 AM
Note 2561.6, 26-Jul-1995
Gladstone:
-----------
I had wished I had it when I had to drive 1 hr. (each way) to press the
Hold Screen button on the console monitor because someone
accidentally leaned on it.
Note 2561.7, 27-Jul-1995
Nagy: Experience with PCM under Digital UNIX?
---------------------------------------------
Does any have experience with Polycenter Console Manager running
under OSF/1 -- er, Digital UNIX, that is.
I'd like to replace our outdated VCS systems (VS 3200s) with an
Alpha and PCM but I won't be allowed to run VMS on the Alpha.
Note, we have been using a number of VCS systems for years and
are very happy with the product.
Note 2561.8, 27-Jul-1995
Aldridge: Comments on Polycenter Console Manager (PCM)
------------------------------------------------------
re: .0 Polycenter Console Manager
We switched from VCS/PCM on a VAXstation to PCM on an OSF/1
workstation. We did this because we wanted PCM to be on the
same platform as our Polycenter NetView -- which never materialized
for OpenVMS VAX.
I don't know about PCM V1.5 for the VAX, but PCM V1.5 for OSF?1
is quite buggy. The OSF/1 workstation crashes which means that we lose
console logging for that time period. We think the review/extract
command does not obey before/since parameters correctly.
It is a big time saver as far as the off-hours commuting goes!
Also I cannot imagine using paper console logs for 10 different
systems!
The DECUServe Journal September, 1995 Page 43
POLYCENTER Console Manager
Unfortunately Digital has stopped giving away with the product the
"filters" for the various operating systems/hardware platforms. There
are some filters available via DSNlink but for the most part these are
not for the most recent systems. That means you'll have to find nice
DECUServe participants or review console messages and create your own
filters.
As an added value I would recommend finding a package such as RamPage
(OpenVMS package) to use along with your PCM system. That way with
just a little effort you can receive a "page" -- either numeric or
full-text -- when a major problem occurs. (After a few of these you
will be sure to resolve the problem, not just ignore the problem!)
As far as the bugs are concerned, we hope some of these will go away
with the newer version of PCM and also Digital Unix (OSF/1) V3.2.
On the OpenVMS platforms I know that the initial version of PCM was
just the next version of "VCS." But I don't know about the follow-on
versions of PCM -- were they re-writes or did they build upon the VCS
base?
Note 2561.9, 27-Jul-1995
Aldridge: More on PCM for Unix
------------------------------
re: .7 -- Nagy
I guess I posted at the same time as you did.
A few more notes about the OSF/1 version of PCM --
1 - All of our console connections are via terminal servers
(typical configuration). We had been using an Emulex server
with the VAX VCS, but had to switch to a DECserver because the
LAT on OSF/1 would not automatically connect to the console ports.
(We could get around it by manually creating the LTA ports [ala VCS]
but this is not how PCM is normally used.)
2 - I would encourage you to get a loaner PAK for PCM and compare it
side by side with VCS. I'm sure you will find things you don't like
but there is something to be said to pick the product that is receiving
more development/support from the vendor.
Note 2561.10, 27-Jul-1995
Zorn: See Freeware CD
---------------------
> Unfortunately Digital has stopped giving away with the product the
> "filters" for the various operating systems/hardware platforms.
I just purchased PCM for VMS/AXP and am in the process of setting it
up. Filters are not included, but a large number of the are included
The DECUServe Journal September, 1995 Page 44
POLYCENTER Console Manager
on the Freeware CD.
I currently manage two data centers, one which is unmanned, the other
which is manned only during business hours. I live an hour from one and
2 hours from the other. PCM has been nice for those times when there
have been site problems. (Now if they would add in failover capability
it would be better)
Note 2561.11, 27-Jul-1995
Clegg: StingRays, AC, and UPS?
------------------------------
Re .1:
You mentioned you have an MTI StingRay. I've got the StingRay II system.
Is that manageable by PCM and if so is it worth it? In the past year and a
half of having my StingRay I've never had to get into it for anything. I
guess that's a good thing. :-) Thanks.
Also, is anyone using PCM to watch/manage any datacenter environmental
systems (AC) or UPS systems?
Note 2561.12, 4-Aug-1995
Cristofero: Remote Support and Stingrays
----------------------------------------
1. I have four systems (4105's ) in four separate states. Using reverse
LAT connections I can have OPA0 functions at my home /office or other
site by just logging into the VCS/PCM node.
We also have a VXT terminal with all the pretty ICON's .
2. Stingrays: I have all my Stingrays (II's II+'s and III's) monitored.
It helped us certify SCSI bus reset's the SRI's were getting from the
I/O being pushed by the our 7000's. Now we have SRIII's.
Note 2561.13, 4-Aug-1995
Clegg: Network -v- RS232C
-------------------------
> 1. I have four systems (4105's ) in four separate states. Using reverse
> LAT connections I can have OPA0 functions at my home /office or other
> site by just logging into the VCS/PCM node.
All the systems I'm targeting for PCM monitoring are local and are in the same
data center. I find myself experiencing some reluctance in configuring PCM to
communicate over the network, reverse lat, or whatever rather than via good
ole RS232C - simply because if the network is down then I can't get to my
systems. All of the VAXen except one can stay up and functioning without the
network - if PCM is dependent on the network then it would be down too. I
don't think that's a good idea. Does anyone else have any ideas or
The DECUServe Journal September, 1995 Page 45
POLYCENTER Console Manager
suggestions or experiences in this area? Thanks.
Note 2561.14, 4-Aug-1995
Gladstone:
-----------
We have had the same argument and I said that if the net is down then
the users can't get to the systems either. We keep a VT terminal in the
data center with a long cord so we can connect directly to a system if
needed.
Note 2561.15, 8-Aug-1995
Clegg: VAXstation 4000/60
-------------------------
I'm now considering a used VAXstation 4000/60 with 1 GB of disk and 64 MB of
memory. I'll use it to watch my 5 VAXen and new RS/6000 AIX system (either a
J30 or an R30 - high-end will be two R30's!)
I've considered the implications of going with network or RS232C connections
and the network is definitely the easiest/cheapest/most flexible/etc. I've
decided that my concerns over accessibility when the network is down are
relatively minor. There's been no time in the last three years where I *-HAD-*
to get into the VAXen when the network was down. Typically during network
down times I am present and actively involved. Also, as a 7/24 shop we got a
little redundancy built in; we can take down portions of the network and leave
other portions up and going. Typically this meets all our needs for network
hardware/software modifications and repair. Total network downtime is very
rare and usually very short lived. The segments containing the VAXen haven't
been down in more than a year - except for the total power outage during a
storm one weekend. (The UPS system is coming in at the end of September.)
I do like the idea of keeping a VT and cable handy for those potential times
when I might have to use it.
Thanks for all your input. Hopefully all this will go through with the powers
that be and I'll get my PCM soon.
Slow System Startup
---- ------ -------
Abstract:
The original poster had noticed slow system startup times under
OpenVMS V6.2, and traced the cause to an INSTALL in the startup
process. A workaround involved frobbing DEADLOCKWAIT, but the
ultimate answer turned out to be a pointer to a DSNlink patch.
The DECUServe Journal September, 1995 Page 46
Slow System Startup
Participants:
Lee Gleason, Michael Mazzoni, Chuck McMichael.
Conference: VMS
Note 2556.0, 17-Jul-1995
Mazzoni: VMS v6.2 startup problem
---------------------------------
I'm having a problem starting up a VMS v6.2 system. After doing the v6.1-6.2
upgrade, I noticed that the time it takes to start the system tripled. At
startup, I put all the things that get done into a command file and submit it to
batch, e.g.:
@SYS$STARTUP:start_product_1.com
@SYS$STARTUP:start_product_2.com
@SYS$STARTUP:start_product_3.com
@SYS$STARTUP:start_product_4.com
etc.
The batch job runs under the SYSTEM account. Now with v6.2, about 1/3 of the
way through the process things slow down severely. If I do a $ SHOW
PROCESS/CONTINUOUS on the batch process, I see that every time the INSTALL
utility runs, it sits in an LEF state for 10-20 seconds. Before it hits the
speed bump, an INSTALL takes 1-2 seconds.
I think I've run out of a resource, but I'm having difficulty finding it. I've
compared the v6.1 and v6.2 SYSGEN parameters and user quotas, and the only
parameter/quota that's smaller on v6.2 is SCSBUFFCNT.
Does anyone have any ideas what could be making INSTALL go into its long wait?
Note 2556.1, 17-Jul-1995
Gleason: Let me know when you figure it out...
----------------------------------------------
Nope, but, I have the exact same problem under VMS V5.5, so it's not V6.2
specific...