[comp.dcom.modems] SLIP compression...

jparnas@larouch.UUCP (Jacob Parnas) (07/03/89)

I recently got SLIP going (4.3BSD), and while pleased with its function,
I still long for better performance even with V.32 modems.  The obvious
thing to do is add compression.  I knnow there should be a version of SLIP
with header compression coming soon, but I am wondering if anybody has
thought about compressing the contents of the packets as well as the headers.

Has anyone out there implemented or thought about implemeting this?  If
you have, I'd appreciate it if you would drop me a note.

Also, does anybody know if a modem doing MNP level 9 compression in hardware
would help?  (such as the Microcom QX/V.32c)

Also, if one were  add compression to SLIP, where should it be added?
Would it be best to add it to if_sl.c?  

Thanks for any thoughts.

------------------------------------------------------------------------------
| Jacob M. Parnas                    | DISCLAIMER: The above message is from |
| IBM Thomas J. Watson Research Ctr. | me and is not from my employer.  IBM  |
| Arpanet: jparnas@ibm.com           | might completely disagree with me.    |
| Bitnet: jparnas@yktvmx.bitnet      \---------------------------------------|
| Home: ..!uunet!bywater!acheron!larouch!jparnas | Phone: (914) 945-1635     |
------------------------------------------------------------------------------

jqj@oregon.uoregon.edu (JQ Johnson) (07/09/89)

In article <340@larouch.UUCP>, jparnas@larouch.UUCP (Jacob Parnas) writes:
> I recently got SLIP going (4.3BSD), and while pleased with its function,
> I still long for better performance even with V.32 modems.  The obvious
> thing to do is add compression.

One possible place to put compression is in the modem itself.  For example,
when I got my TB T2500s to set up a SLIP connection the package included
the announcement that Telebit planned to (but didn't yet) support MNP.5
in V.32 mode.  Does anyone have any experience with SLIP performance using
the standard hardware-based compression algorithms (MNP.5, V.42 or whatever)?
It clearly won't do as good a job on IP headers as VJ's algorithm, but it
also doesn't require hacking SLIP code for which you might only have
binaries.

vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (07/10/89)

In article <5108@oregon.uoregon.edu>, jqj@oregon.uoregon.edu (JQ Johnson) writes:
> One possible place to put compression is in the modem itself...[use Telebit's
> normal compression]

After hacking a form of TCP compression into SLIP, which does nothing for
stuff like ICMP, I decided to see what the compression built into a TB+
would do for ping.   The packets sent by ping contain almost no
"information."  The default 56 bytes generated by the 4.3BSD ping is a
simple count.  The IP & ICMP headers do not change very much.  Naively, I
thought a TB+ might compress such a slow, repeative byte stream into
something which would fit into "micro packets."  I had no such luck.
Why?  Has anyone had better luck?

Vernon Schryver
Silicon Graphics
vjs@sgi.com