Path: utzoo!attcan!uunet!decwrl!ucbvax!UM.CC.UMICH.EDU!Gavin_Eadie
From: Gavin_Eadie@UM.CC.UMICH.EDU
Newsgroups: comp.protocols.appletalk
Subject: Re:  Re: mactcp and SLIP
Message-ID: <6702980@um.cc.umich.edu>
Date: 21 Aug 90 12:28:26 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 22

> My understanding (John V., care to clarify this?) is that the timeout values
> used by MacTCP are too quick for the (relatively) extremely slow speeds at
> which SLIP and PPP transfer data.
 
This may be true for SLIP in which the initial requests for an IP
address are pumped out rather fast - too fast for one frame to have
been completely sent before the next is started - but it is not true
for SLFP (an old MIT variant we use in the Merit network) because
SLFP delivers an IP address to the Mac immediately, with no poll. We
have a working SLFP LAP for MacTCP which we wrote early in the year
and we have been working on a PPP implementation recently.
 
There are some problem though. It's impossible to tell when you can
hang up the phone since the Close calls to the TCP layer are not
passed down to the LAP (not to mention the fact that you could get
UDP data at any time without warning. I believe large packets cause
problems sometimes (probably a timing issue) and there are some
strange aspects to the MacTCP cdef interface that cause multiple
LAPs to be highlighted simultaneously!
 
Gavin Eadie, U of Michigan Information Systems (speaking for others).
Internet: Gavin_Eadie@ummts.cc.umich.edu