Position & proposal

David Gadbois gadbois@cyc.com
Wed, 4 Jun 1997 23:59:49 -0500 (CDT)

   From: "ET" <emergent@eval-apply.com>
   Date: Wed, 4 Jun 1997 19:24:56 -0400

   I wouldn't suggest hot patching a sever under a heavy
   load, but that it would be nice to deploy a server under

Actually, hot patching is SOP under Genera.  Because of the reflective
nature of the system (both statically and dynamically) it is easy to
locate and deal with all the interdependencies that may come up when
you make low-level changes on the fly.

One case I dealt with was a bunch of flaky Macintoshes on the network
that would occasionally do a SYN flood that would cause
denial-of-service problems for the lispms.  It was a simple matter to
locate the code that dealt with the MAC/IP boundary, modify it drop
the malformed packets on the floor in such cases, and make a patch
with the fix.  The automatic patch loading mechanism took care of
distributing the patch to all the running lispms.  The users never
even noticed.

And it all works the same at the application level, too.  For years at
Cyc, we have been hacking on the system, making major modifications
and fixing bugs, all the while the Lisp users were using it to get
work done.  Nowadays, the folks using the C versions have to wait for
the weekly build to get updates.  It sure does hurt productivity when
you have to wait days rather than minutes for a problem to be

--David Gadbois