[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

FTP Problems (was: Symbolics terminal emulator works lousy)



You write:
    How about Logout?

Essentially, when everyone using the Lispm plays by the Rules (Logout,
Reset Network, and all the other things you all have suggested),
things go ok.  Our other users, however, seem to manage with little
difficulty to get the Lispm into a state where these things are no
longer possible (the machine will stop itself, or drop irreversibly
into the cold load stream, or something . . . maybe they're just
panicking, I don't know - I just get the complaints).  Essentially,
playing by the Rules is not an option here.

It just is, and should remain, possible to get the Lispm into a state
where it "cannot fulfill its part of the TCP bargain".  Forbus is
right, adding all that protection is like making a race horse drag
rocks.  In spite of the fact that it is as big as a Frigidaire, the
Lispm is still the moral equivalent of a PC, which means you have to
use the reset button, perhaps a lot.  That's just what I want.

However, by consequence, given the nature of the TCP protocol and of
the Lispm, THAT THESE TWO ARE FUNDAMENTALLY NOT MEANT FOR EACH OTHER.
Maybe that's why Chaos works better.  Our rep tells me that Chaos on
Unix is no longer supported.  Is this *really* true?

When I used these machines in the US, I did not have to system hack
--- we had wonderful support staff.  As I recall, we first ran Chaos,
and then switched to TCP in about 84, but none of these problems
popped up with the Unix machines there.  So when I asked the initial
question, I had a preconception that there was a solution around
somewhere, but that I just couldn't find it.

The answer I seem to be getting is, though, that there's just no way
to make this run smoothly.  Is this really true?  In that case I go
back to
1. my hourly cron shells.  I don't like this, because:
   a. from the unix utility side, its hard to tell a connection being
      used from a dead one.  I consequently run the constant risk of
      getting an angry user in my office complaining about being
      salvoed in the middle of a file transfer or something (people
      are best at complaining here). 
   b. I haven't the foggiest notion yet about how to deal with VMS.
   c. The next time they hang something bizarre on the network, the
      problem starts over again.
2. another solution I thought of later, which would be what I call a 
   "tabula rasa" protocol.  That is, when you boot the machine, it
   comes up, and then invokes a broadcast service on the net that says
   essentially "Hi, I'm alive, and drop everything you have had going
   with me up to now".  I like this one, but it would require a lot of
   extra hacking, and points 1b and 1c still apply.  Better would be
   something that left the non-Lispm side software alone.

   However, in view of what I said above, it seems that whatever it
   is, it must run on the non-Lispm side.
 
Surely there's someone out there who has solved this problem more
satisfactorily?  ceb