From: CRDGW2::CRDGW2::MRGATE::"SMTP::CRVAX.SRI.COM::RELAY-INFO-VAX" 27-JUL-1990 05:16:39.52 To: MRGATE::"ARISIA::EVERHART" CC: Subj: re: DECnet over TCP/IP Received: by crdgw1.ge.com (5.57/GE 1.70) id AA06836; Wed, 25 Jul 90 09:26:12 EDT Received: From UNIX.SRI.COM by CRVAX.SRI.COM with TCP; Wed, 25 JUL 90 06:04:47 PDT Received: from yale!LRW.COM!lrw by unix.sri.com (4.1/SMI-4.0) id AA08073; Wed, 25 Jul 90 06:01:58 PDT Received: by harvard.harvard.edu (5.54/a0.25) (for sri-unix!KL.SRI.COM!INFO-VAX@husc6) id AA01892; Wed, 25 Jul 90 07:36:30 EDT Received: from lrw.UUCP by BULLDOG.CS.YALE.EDU via UUCP; Wed, 25 Jul 90 07:14:48 EDT Message-Id: <9007251114.AA15378@BULLDOG.CS.YALE.EDU> Received: by lrw.UUCP (DECUS UUCP w/Smail); Wed, 25 Jul 90 07:03:28 EDT Date: Wed, 25 Jul 90 07:03:28 EDT From: Jerry Leichter (LEICHTER-JERRY@CS.YALE.EDU) To: INFO-VAX@KL.sri.com Subject: re: DECnet over TCP/IP X-Vms-Mail-To: IN%"PYM@QUCDNSUR.bitnet",INFOVAX Our Campus Computing Services group (traditionally an IBM shop, now supporting both both IBM and Sun) is determining policy for the use of the campus-wide network (Ethernet over a fibre backbone). There is also a campus-wide DECnet with just under 50 VMS nodes (departmental systems unsupported by Computing Services) - which has grown over the past 5 years and is presently implemented on asynchronous 9600 baud lines (with two of the nodes having synchronous links to the IBM mainframe for Bitnet/Netnorth connections). The original promise was that the new campus-wide Ethernet would carry any 802.3-compliant protocol and that, although TCP/IP would be the supported protocol, DECnet could co-exist with it, thus allowing us to change from asynchronous DECnet. Now, unfortunately, policies are "evolving" - the powers-that-be wish to keep the network "pure" and adhere to "standards", avoiding "proprietary protocols" (the most OSI-compliant protocol does not fit this description?). In other words, DECnet will be banned, and TCP/IP will be the only protocol allowed. The proposed solution for DECnet users is to implement DECnet over IP using Multinet. While this is certainly feasible, we are concerned that the overhead will be crippling. Although network traffic will be a little greater with DECnet over IP compared with unadulterated DECnet, we are most worried about the CPU resources required for stripping the IP packets and handing the contents over to DECnet. Like any university site, we have limited computing resources; many of our VAX nodes are 750's or uVAX-II's (and are serving as boot members for LAVC's as well as compute-intensive interactive and batch use). What will be the impact on these systems? Would we be better to stay with our present 9600 baud asynchronous DECnet? What cogent arguments could we use to convince the powers-that-be to allow unadulterated DECnet over the campus ethernet? I would be extremely grateful for comments on this situation (even flames are welcome!) and any advice that the Info-VAX readership may have for us. My, the "standards" mania has really spread. Or should I say standards "mantra". A more severe case of ACI syndrome (ano-cranial introversion) I have not heard in quite some time. It's difficult to know how to respond to this kind of idiocy, but let's see what we can come up with: - DECnet is NOT a proprietary protocol. Its specifications have been published for quite some time, and there are several third- party implementations for a wide variety of machines. (In fact, it may very well be that there are more INDEPENDENT DECnet implementations than there are INDEPENDENT TCP/IP implementations: Almost all TCP/IP implementations are ports of some version of the BSD one.) - DECnet in Phase V - not very far out - will comply with OSI stan- dards. These standards will form the INTERNATIONAL standards for networking in years to come. Anyone committing themselves to a TCP/IP network today must realize that they will probably be converting to OSI some time in the future - perhaps not for a number of years, but it WILL happen. (The technical merits of OSI and TCP/IP are not the issue here, and I'm not inte- rested in debating them. For various NON-technical reasons, it's clear that this will happen. Even the US DoD, which sponsored TCP/IP to begin with, has committed to OSI.) - Ethernet is a standard. The Ethernet STANDARD provides for multiple protocols on the same Ether. That's an intentional design feature. There is absolutely nothing non-standard about running multiple protocols on an Ethernet. - Conversely, DECnet over IP is NOT a standard of any kind; it's a vendor-specific extension. Then again, the TCP/IP world (writ large) is full of extensions: VMTP, multicast IP, etc. There's nothing at all wrong with this: TCP/IP is not a solution to all networking problems (nor is DECnet or OSI). It's a sign of vitality that people are attacking new problems with new solutions. But anyone who thinks they can specify "TCP/IP only" or even "standards only" has, well, ACI Syndrome. - Here's one that should get to some people: You're a Canadian site. Point out that TCP/IP is an US standard, NOT an international one. The INTERNATIONAL standard is OSI, which is what DECnet will shortly be. (Check my facts on this one before using. I'm pretty sure they are correct - I can't imagine ISO being willing to entertain an alternative network standard after all the work on OSI - but you never know.) The only cogent argument I know of for limiting an Ethernet to "IP only" is that you can then use IP-level gateways between segments, rather than lower- level bridges. However, there are very few configurations in which this gains you any advantages - usually it's a more expensive way to get less perfor- mance. -- Jerry