TCP/IP on XENIX 386 && 3c501

[Russ Nelson]

In article <1990Nov21.030502.18951 at simasd.uucp> jcurtis at simasd.uucp (Jay Curtis) writes:

   In article <NELSON.90Nov20164937 at image.clarkson.edu> nelson at clutx.clarkson.edu (aka NELSON at CLUTX.BITNET) writes:
   >Be sure to use the UNDOCUMENTED onepacket switch on ifconfig if you're
   >using a 3c501.  The symptoms of failing to do this are that bulk
   >transfers to systems running SunOS proceed exponentially slower.
   >The most I could ever transfer before onepacket'ing was 50K.
   >
   >I cannot emphasize this strongly enough, so I won't bother.  Just
   >remember this if you ever put a 3c501 in a SCO Xenix box, and maybe
   >Unix as well...

   Okay, I'll bite.  What is this "UNDOCUMENTED onepacket switch" ?  Would
   you care to enlighten those of of who are cursed with no alternative to
   using the 3c501?

Don't use the 3c501 if you can possibly help it.  Beg, borrow, steal,
whine, do whatever it takes the avoid it.  Any of the 3c503, ni5210,
or wd8003e would be much, much better.  You see, the 3c501 has only *one*
packet buffer, so you cannot send it back-to-back packets, *plus* it uses
that buffer for sending packets, so while you're preparing to send a packet,
you're deaf to the network.

But if you really, really have to use one, then do a:

	ifconfig 3comA0 onepacket

in addition to the other uses of ifconfig.  It doesn't hurt to run
ifconfig several times, so you could add the above line verbatim to
/etc/tcp.  I'm guessing that it closes the TCP window so that only one
packet can be sent at a time.  You can imagine what that does to your
throughput.  Of course, NOT setting onepacket is even worse, because it
means guaranteed dropped packets.

--
--russ (nelson at clutx [.bitnet | .clarkson.edu])  FAX 315-268-7600
It's better to get mugged than to live a life of fear -- Freeman Dyson
I joined the League for Programming Freedom, and I hope you'll join too.