On Sun, 6 Apr 2003, Christos Zoulas wrote:

> In article <Pine.NEB.4.44.0304061318460.6413-100000@tautology.immanent.net>,
> Frederick Bruckman <fredb@immanent.net> wrote:
> >On Sat, 5 Apr 2003, Iain Hibbert wrote:
> >
> >>    I wrote a decompressor for PPP using the Stac LZS algorithm, its at
> >> <http://homepages.rya-online.net/plunky/lzs.html> and has been working
> >> well for me for some time without any problems..
> >
> >That's pretty cool. Even with one history buffer, you should be able
> >to get +14% for large plain text/html/uuencoded downloads

more than that - my connection is a mobile phone link that sometimes
reaches 1200 bytes per second according to pppstats, but links reports up
to 3K/s on normal web pages (some computer generated ones are full of crap
and compress even better), and I got 10K/s on a highly compressible file
with ncftp.

Extra buffers would only be useful if the compressor could identify
different data streams since you reference one history per packet. A
larger buffer might make a difference sometimes but the buffer size is
fixed in the rfc..

> >Would you mind filing a problem report, so at least it stays visible
> >until someone picks it up?
>
> He should send-pr it.

Alas there is patent nonsense involved, so I doubt NetBSD can incorporate
this (sigh). I'm happy to send-pr it though, if it would stay in the
database..?

iain