Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!sdd.hp.com!decwrl!sgi!vjs@rhyolite.wpd.sgi.com
From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
Newsgroups: comp.sys.sgi
Subject: Re: Streams under IRIX?
Summary: yeah, we've STREAMS
Message-ID: <94144@sgi.sgi.com>
Date: 28 Mar 91 19:05:31 GMT
References: <1991Mar26.215809.19943@ux1.cso.uiuc.edu> <93971@sgi.sgi.com> <3542@lamont.ldgo.columbia.edu>
Sender: guest@sgi.sgi.com
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 38

In article <3542@lamont.ldgo.columbia.edu>, dale@lamont.ldgo.columbia.edu (dale chayes) writes:
> > 
> > Well, STREAMS won't help much with Silicon Graphics TCP/IP since IRIX is
> > currently and in the forthcoming release firmly socketed.
> ...
> 	So, is Streams there or not? (in 3.3xx)


Silicon Graphics has been shipping SVR3 STREAMS for a long time.  I think
they were in 2.5/3.5, and I know they were in 3.6 (for the IRIS 3000
series).  I put them in for reasons of varying validity, including the
notion that it was the easiest way to implement PTY's, and the fact that
the STREAMS tty's on the 3000 were 12 times faster than the SVR0 clists.
As far as I know, we were shipping STREAMS TTY's before anyone else,
including AT&T and the solar guys.  This is nothing to brag about, because
the hacks I came up with to make STREAMS tty's pass SVID are amazingly
close to the independently invented solar hacks, but they are not
identical.  Our line discipline module is also amazingly similar on the
outside, but also not identical.  Someday we may take the SVR4 punishment
for being early.

We only changed the SVR3 STREAMS head and utility code to work in our
symmetric-MP kernel, to pass the AT&T SVID, and to work in a lot of places
where AT&T thought "x->foo |= bit" is a non-interruptable operation on all
hardware.

One of the reasons for doing STREAMS ttys was to help decide how to get
good TCP/IP.  Ultimately, we chose to do a straight 4.3BSD to SVR3
shoehorn.  This means we have STREAMS for tty's and for any drivers
customers or third parties write, but we have sockets for networking.

Thus, if you want TCP on an IRIS, you're stuck with sockets, for at least
the next release or so.  It could be worse; check the ttcp performance
numbers on localhost when you get 4.0.



Vernon Schryver,   vjs@sgi.com