🚀 go-pugleaf

In article <1992Oct29.115804.14105@decvax.dec.com>, evans@zk3.dec.com (Marc Evans) writes:
| 
| In article <1992Oct27.104816.3673@cucs5.cs.cuhk.hk>, shlam@eng.ie.cuhk.hk (Alan S H Lam) writes:
| |> I am trying to install SLIP at DEC 3100 running Ultrix 4.2a.
| 
| Without hardware flow control (which the DS3100 does not have), you may as well
| not bother trying to get SLIP running. It will not work at any degree of
| satasfaction.

I don't know how much worse Ultrix is than VMS+MultiNet in this regard, or
how much worse a DS3100's serial port is than a VS3100's, but I have found 
that is NOT necessarily the case that you need hardware flow control effectively
to run SLIP.

We have run SLIP at 38.4Kbps on a MicroVAX 3400 CXY08 and at 9.6Kbps on
a VS3100 port, with no hardware flow control in either case, and found
in both cases that performance was quite good (relative to the line speed),
and that there were virtually no dropped packets.  For example, we found
consistent FTP throughput on the 38.4Kbps SLIP link of over 32Kbps.

One factor may be that TCP has its own notion of flow control and seems
to do a decent job of not overrunning the line.  (It may be that using NFS
over UDP with 8192-byte blocks would NOT work so well!)

It *is* a good idea to turn down the MTU on the serial port to, say, 296 bytes,
and to turn on VJ header compression.

Aaron

Aaron Leonard (AL104), <Leonard@Arizona.EDU>
University of Arizona Network Operations, Tucson AZ 85721
  "It's not a bug, it's a form of flow control."
  - Jerry Leichter on why crash-prone Unix is a suitable
    platform for NSFNET core routers