lew@t4test.UUCP (Lew Mullen) (05/29/85)
Is there anyone running Eunice yet under VMS 4.x ? (like VMS 4.1?) We are going to have to do it soon, and would like to think we aren't the first site to anticipate a problem. First thing is "it's different". (like 25 character filenames, or so) Second is ... we will be running a 2-cpu VMS cluster ! Has anybody got a sales pitch from TWG on this ? Anything ? Say, that reminds me, what ever happened to that newsletter they started ? I got a couple of issues, then they stopped coming. -- t4test!lew ( Lew Mullen @ Intel/Santa Clara )
lew@t4test.UUCP (Lew Mullen) (06/25/85)
Following is a summary of responses I got from this inquiry about Eunice under VMS 4.1: Is there anyone running Eunice yet under VMS 4.x ? (like VMS 4.1?) Say, that reminds me, what ever happened to that newsletter TWG started ? ----------------------- > From dual!mordor!seismo!sri-iu!kashtan Sat Jun 1 01:57:23 1985 > (David Kashtan) > > I did the VMS 4.0 work for Wollongong under their agreement with SRI. The > VMS 4.x version is currently running here at SRI and at Stanford. The biggest > change is that hashed filenames are almost NEVER necessary -- there is a new > encoding scheme for NON-VMS filenames. There is a utility in /etc that will > fix up all your old filenames (/etc/cvtfnames \[*...]). It is also the case > that the EUNICE logical names have changed to be fully compatible with VMS > rooted directory specs. (and search paths). So now /etc is defined as > ETC ==> DEV:[XETC.] > rather than > ETC ==> DEV:[XETC] > > This allows you to use search paths for clusters -- i.e. there is a cluster > wide ETC with a node specific ETC for each node in the cluster (pretty much > the same as the SYS$COMMON stuff). The ONLY time you notice this difference > in the logical names is when you try to access something in /etc with a VMS > filespec -- instead of ETC:FILENAME.EXT it is now ETC:[000000]FILENAME.EXT > > Anything else I can tell you? > > P.S. The Stanford site is running a 4 machine cluster with almost ALL > the EUNICE files in common on the cluster. ----------------------- > From: hplabs!epicen!jbuck > Joe Buck Entropic Processing, Inc. (epi) > > We received the distribution for VMS 4.0. Don't install it! They screwed > up execution of privileged (installed) images, so you can't run commands > like ps or mail without privilege. They are frantically working on a fix. > > > Editor's note: TWG has this fixed ... It was some of those > > VMS "magic underscores" acting up again. - lew > > The release notes said it won't work properly on clusters. > > > Stanford got it to work OK (see Kashtan, above) - lew > > (Maybe the Stanford people didn't read the release notes) > > I assume the hashing scheme is different, since VMS filenames are much > longer. However, Eunice is still a 4.1, not a 4.2bsd emulation, so > I assume you won't be able to access VMS files with names longer than > 14 characters. > > > Care to comment on this one, anyone ? - lew ----------------------- > From: packard!topaz!hedrick (Chuck Hedrick) > > We are running TWG's TCP under 4.0. Since it runs under Eunice, Eunice > must be working. ----------------------- > From hplabs!decvax!cwruecmp!diamant > (John Diamant) Case Western Reserve University CSNet: diamant@Case > > Here at Case, we have a VAX 11-782 which we just brought up with VMS 4.1. > We are in the process of getting EUNICE to run under it. The stupid > HSH... filenames are gone as VMS can now handle "." files properly. They > have a program which is supposed to convert the old format to the new one > (called cvtfnames or something) which doesn't seem to work quite as > advertised. Also, when EUNICE is entered, some logicals are defined, which > aren't properly removed when it exists. As a result, when it is reentered, > it says "can't find /etc/passwd entry." It is a simple matter to change > the cshell.com file to delete the logical manually (only the one called > EUN_5C_ENV is important). We are thinking about breaking our two cpu's > apart and making it into a cluster, so would be interested in your experiences > in this area. Other than the two minor points described above, there > haven't been any major problems with bringing up the new version. ----------------------- Last, but not least, I had an informal talk with Jeff Wheelock of TWG, whose staff will be supporting Eunice now that Michael Bourke has left. The best thing I saw was that TWG has recompiled a large part of the sources, which means among other things they HAVE some of the sources to look at when you have a problem. Hey, David, how about pointing them at the UUCP sources ? Filename hashing has been dramatically improved, with no obvious need for those pairs of *.HSH *.HSN files anymore. YEAH ! They are also adding code to limit the number of users to the number you paid for on your license; supposedly required by AT&T. Every program will check the number of users before it lets you run it. (I hope this doesn't slow it down TOOO bad) I asked our salesman for a cluster-wide *nix-binary/Eunice license, so far no progress. Seems to be another problem with AT&T. p.s. The newsletter is quarterly. One should be due any minute. ----- The opinions expressed or implied by this document are mine alone and not necessarily shared by anyone else. -- t4test!lew ( Lew Mullen @ Intel/Santa Clara )