 Things to do for BRANCHES:  K 45. It may be possible to do something about the volume switching on backup L reads to automate the replys.  If we recognize that a backup read request isO outstanding on a drive and the operator then causes a volume to be switched out J and then switched back in.  That could trigger the reply.  That needs some thinking about.   N 48. Think about wrapping the inventory table structure in a class and stickingM all the miscellaneous operations there.  This will cut down a LOT on compiles 5 and make it easier to modify the structure correctly.   I 51. Write a scsi error to text processor for all the libraries to produce K meaningful SCSI errors.  This would have to be done on a library by library N basis since there is some room for variation in text and interpretation in the specification.  K 52. Handle the various application specific mount requests via the means of K a callable image.  The logical name gets defined per box and can be used to J do just about anything.  Great for implementing the DFS or NFS enable code after mounting.   9 54. Still need to deal with the following license issues:    	1. multiple licenses per node. " 	2. multiple use license per node.  J 58. Recast most of the find operations in term of the foreach* operations.  O 67. Write a SCSI pass through driver that allows [at least] an abort message to  be sent.  M 69. API to ask questions about relationships between boxes and devices, e.g., O given a box, tell me how many devices there are, what they are, etc.  Likewise, E given a device, tell me if it is controlled by a box, which box, etc.   P 96. I need to do a better job of encapsulating the SCSI commands that get fed toO various boxes.  Right now, variations in SCSI commands (like the need to assert O vendor specific bits) have a HUGE impact on the code.  I'll need to handle this O all as part of the effort to get the UNLOAD functionality back in place for the  SA100 product.  See item # 92.  I 97.  This relates to item 69.  If the server goes away while the Branches M interface is doing a mount or dismount, the interface will hang.  This can be P fixed by asking the database which box is related to the drive and then checking for THAT server being alive.  I 103. Load stall can be implemented in ways other than by timeouts!!!  For L libraries implemented as different LUNs off of the same SCSI id, I believe IM can always ask for status via test unit ready and when the unit stops showing L busy, release the appropriate mechanisms.  This will pick up performance forL things like DEC and Conners stacker.  It won't help the libra, but you can't have everything.  J 105. Since the WangDAT was derived from the SA100 and the SA100 doesn't doN unloads properly, the WangDAT now doesn't do them either.  This was fixed withO the THZ27.  If I rearrange the inheritance to be from the thz27 to the wangdat, M I think that solves all the problems.  This will be easier than introducing a # new class for a WangDAT drive slot.   K 119. ADIC unloadStall can use inquiry to find out if the cartridge has been  ejected.  M 120. Python loadStall can use test unit read to find out if the cartridge has  been inserted.  K 123. There is a state in which volumes are both MOUNTED and DISMOUNTING (it J shows up in show device).  I'm pretty sure this will prevent Branches from7 seeing the dismount request.  I need to check the code.   P Device                  Device           Error    Volume         Free  Trans MntP  Name                   Status           Count     Label        Blocks Count CntP QUICK$MKA200:           Mounted dismount     0     USER2             0     1   0%                         alloc foreign   ! Is the state seen by show device.   M 124. ADIC type libraries don't seem to do the "right" thing with INIT /VERIFY ' /USAGE=BACKUP.  They verify everything.   & 125. Need an allocation function like:           allocate drive!         allocate slot (or volume)   * 126. Need an initialization function like:  D         import /initialize /select=scratch (init any scratch volume)  * To make backup procedures easier to write.  L 127. Need a function like import /magazine /scratch to get a load of scratch volumes into branches.  P 128. We need to support the removable magazines in the Exabyte products better. L We need to implement some kind of import/magazine[=location] capability thatM would let folks put magazines in place and then import/export their contents.   M 147. Add protection information to individual tapes.  This would prevent them M from being used by unauthorized persons.  Do it in two stages:  (1) UIC based 	 (2) ACLs.   M 161. If a mount request for a continuation volume appears for a drive that is L already mounted, then dismount the current volume in the drive and mount the- requested drive.  This must only happen when:    (1) The drive is allocated. L (2) The volume requested is specifically a continuation volume, i.e., eitherK xxxx02 or xxxxnn where the NEW nn is exactly 1 greater than the current nn.    	Done. Ship with 2.4A.  H 	There is still an issue with BRANCHES DISMOUNT which appears to get theH 	"wrong" volume name from getdvi.  This isn't as much of a problem as itF 	would appear since the dismount still occurs and the dismount messageI 	provided by VMS gets the right label.  I can get this right by recording C 	the volume label with the original mount request (always) and then I 	checking the volume label the the B DISMOUNT against the original volume  	label.   / 162. Figure out what to do with cleaning tapes.   K For hardware with cleaning cartridge slots, define them automatically.  Add M capability to reserve slot for cleaning cartridge use (in configuration file:    CLEANING-CARTRIDGE	slotNumber   
 Allow only 1.   M Import/magazine ignores cleaning cartridge.  All other operations work on it. " Import won't pick it if it has to.  L 164. We can handle the relative volume grabbing a scratch volume by taking aO look at the read only state after the mount has occurred.  If the state is read N only, then we don't mount scratch tapes.  Otherwise we assume that we can grab7 any scratch tape and force an initialize, if necessary.   3 173. Catch the transport not working in mode sense.   $ 186. HyperTape Integration examples.  & 187. Documentation of Tapesys support.   190. Put the basic operations:   	initialize element status 	read element status 	mailbox table
 	drivetodrive  	drivetomailbox  	drivetovolume 	mailboxtodrive  	mailboxtovolume 	volumetodrive 	volumetomailbox  P Into the odl object.  Start writing everything in terms of those operations.  ItN took me over 3 years to realize this!!!  Boy am I dumb!!!  This should make itG possible (with the database) to completely encapsulate ALL of the basic M operations for Branches into 2 classes, odl and odldatabase.  Everything else 8 should be writeable in terms of the movement primitives.  M After this is done, a MAJOR pass through Branches must be made to rip out all P redundant classes that simply act as renamings of other classes.  Then we retireJ the old CMS and reference library and start from scratch.  This version ofN branches will be version 3.0.  Technically I could get rid of the slot objectsM entirely, but that would be WAY too big a change.  What I'll do instead is to P move the algorithms into the basic slot objects and shoot all others.  This will" take about a month to do AND test.  * 194. Add new member functions of the form:   	load(devicename, &drive)   9 which use the specified drive and DON'T create a new one.   . 195.  Needs an isSame in the Drive Slot class.  O 196. John, correctly, points out that we can handle the problems of backup not  Q doing drive substitution by physically shuffling cartridges around if necessary.  J Maybe a backupMount method? or a mountNoSubstitution method.  Needs to be  thought out a little.   L 197. Import should participate in the "idle" drive unloading if there aren't any drives available.   
 202. Need:   	IMPORT /VID 	EXPORT /VID  N 204. Report of bug in multiple volumeset/tape/multiple tape backup.  Last tapeO in the drive doesn't get updated.  This may happen if the first tape is mounted : using either MOUNT or BRANCHES MOUNT.  Needs to be tested.  $ 208. unload/library doesn't on ADIC.  O 209. Seem to get occasionaly cross talk between servers running on VAX and AXP. O Actually,. we get cross talk between UNLICENSED servers.  I'll need to fix this M at some point, but that requires moving the license check into the individual  function tests.   M 210. IMPORT/MAGAZINE will fail on AXP if drives are all full, but dismounted. / Need to make import unload drives if necessary.   J 223. Allow import to utilize an idle drive.  Add a control to disable this3 behavior, e.g., IMPORT-IDLE-DRIVE 0 to turn it off.   I 224. remove TSA100PHYSICALVOLUME as the generalization of physical volume ' should handle why it needs to be there.   : 229. Remove EX10Iinitelestat which has a bogus handling of the scsi timeout.   