From mboxrd@z Thu Jan  1 00:00:00 1970
To: 9fans@cse.psu.edu
Subject: Re: [9fans] X on Plan 9
From: geoff.9fans@collyer.net
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Message-Id: <20010421074926.7AF5C199DC@mail.cse.psu.edu>
Date: Sat, 21 Apr 2001 03:48:16 -0400
Topicbox-Message-UUID: 880e6d48-eac9-11e9-9e20-41e7f4b1d025

Russ's Palm file servers are definitely preferable to what I've got.
I ported the pilot-xfer stuff (v0.9.3 or 0.9.4) using APE but it
was flakey due to the use of alarm signals to implement protocol time-outs
and possibly due to problems in APE (e.g., select() poisons fds so
that they can't be dupped).  I then ported it to native Plan 9,
cleaning it up in the process.  I think I'd deleted over 35% of the
source by line count when I last looked but more code deserves to be
deleted; there was silly stuff like an emulation of BSD sockets in it.

There are a few remaining embarrassments: my protocol time-out code
consumes processes at a rate that seems excessive to me, and there
remains a window in which a full packet may have been read and yet
it will be discarded if a time-out occurs; the thoughput on the serial
line is lower than seems reasonable, though I haven't compared it with
the Windows Palm desktop hot sync; and the program leaves processes
behind when it exits, which inteferes with later uses of the program.
This last looks like some form of memory corruption that I fear was
in the code when I got it.  I haven't had much time to spend on this
lately; the obvious next step is to try the various debugging hooks
in malloc.

So what I have works, for small values of ``work''.  It's sufficient
for dumping and restoring Palm files (``databases''), so I no longer
need to use Windows nor Linux for those tasks, and that's gratifying.