Re: DNA

[MILLER@VAX.CAM.NBS.GOV]

Reply to
   
  We use DNA between our VAXen and Symbolics.  ...
  All file access is essentially transparent.
  What exactly do you mean when you say "We would like to use the VAX as a file
  server that is capable of handling the Load/Compile system feature of
  Symbolics?" ...  If what you mean is that you want to compile complete systems
  residing on the VAX, then I'd expect as long as your translations were correct,
  the Lispm wouldn't have any trouble finding and loading your files.  Am I
  mistaken?
 
  Len Moskowitz

Unfortunately you are.
     We also use DNA between our VAX & Symbolics. Most things work as they
should, except that it asks for passwords at awkward times. Also, if
many users use the same machine, it insists on remembering passwords and
default pathnames from the previous people, and the VAX gets annoyed.
The password problem can be fixed by advising LOGOUT to do
(DNA:FLUSH-PASSWORD-CACHE).
    There is one thing that I know does NOT work with Load System.
(Assuming the system files etc, all reside on a VAX, and the
translations are set up correctly, of course).  The problem occurs when
a system contains component systems. Suppose system B contains
component system A, and suppose that system A has already been loaded.
When you load system B, it doesn't realize that system A has already
been loaded, so it loads A again. This is a big waste of paging space
especially if you are building worlds.
    But wait, it gets worse! If the system A was patched, when A gets 
reloaded during the loading of B, the system IS smart enough to figure 
that A's patches have been loaded so it doesn't load them again. Now you've
got A loaded ON TOP of its own patches, a curiously inconsistent beast with
low survivability.
   There are two workarounds: 1) store all susceptible systems on a symbolics.
2) create a script which has lots of (LOAD-SYSTEM ... :INCLUDE-COMPONENTS NIL)
forms. You've got to trace through the tree of everything you want loaded and
include ALL component systems IN THE RIGHT ORDER.  Due to limited disk space 
on our Symbolics, I use the latter workaround, but let me tell you it is
painful.
   What I'm curious about is that I reported this about a year ago and when
a colleague of mine ran into the same problem he had to `prove it' to the
support people again.  I know it may be a tricky problem to fix, but at least
admit that it exists, and tell people about it.  In fact: If you are not 
going to fix it soon, tell me, so I'll know it's worth the trouble of 
writing a `system unwinder' to automatically perform workaround 2!!!

  I would like to know if anybody has had similar problems with TCP/IP either
using a SUN or a VAX (especially a VAX w/VMS but TCP/IP software) as a file server?
		Bruce Miller
		MILLER@VAX.CAM.NBS.GOV