Xref: utzoo unix-pc.general:1997 unix-pc.sources:204 comp.sys.att:5102
Path: utzoo!utgpu!attcan!uunet!lll-winken!lll-lcc!ames!mailrus!ukma!david
From: david@ms.uky.edu (David Herron -- One of the vertebrae)
Newsgroups: unix-pc.general,unix-pc.sources,comp.sys.att
Subject: Re: UNIX-PC SLIP sources
Message-ID: <10823@s.ms.uky.edu>
Date: 6 Jan 89 16:58:58 GMT
References: <1157@sw1e.UUCP>
Reply-To: david@ms.uky.edu (David Herron -- One of the vertebrae)
Organization: U of Kentucky, Mathematical Sciences
Lines: 26

The place to get the KA9Q sources is over in the ftp directories
on louie.udel.edu ..

Now, I don't see how it could possibly work.  Someone said this code
had been developed for BSD machines and doesn't currently run on SysV.
This says to me that select()'s were used for polling I/O, since SysV
doesn't have an equivalent facility.  I spent a day last week thinking
over different ways of implementing a select() mechanism and I think I
have a way.

Start off with a shared memory block.  For each file descriptor which
you'd want to be doing select()'s on you fork() off a child process 
to handle that file descriptor.  (Two children maybe?)  The child
has an infinite loop that hangs in a read() call and when that returns
it puts the result into shared memory, and sets some flag in shared
memory saying that the data is there.  Possibly it uses a signal to
let the parent know immediately that some data has arrived?

You need this so that the TCP/IP code is able to be active all the time 
and also let it have a way of suspending itself until something is available
to do.
-- 
<-- David Herron; an MMDF guy                              <david@ms.uky.edu>
<-- ska: David le casse\*'      {rutgers,uunet}!ukma!david, david@UKMA.BITNET
<-- Now I know how Zonker felt when he graduated ...
<--          Stop!  Wait!  I didn't mean to!