[ changed the cc: to be current-users ]

> > i'm not sure that i'd buy that without a measurement to prove it.
> > 
> > 1 PPP line is what?  at most 57.6k/sec?  150k/s?  If done in an
> > optimized manner (e.g. some shared memory region, or something), you'd
> > not have to pass the data back and forth, and, if multiple packets
> > could be done at once, it wouldn't necessarily be _that_ many extra
> > user-kernel switches per second...
> 
> Well, if you stick to running PPP over async lines, sure.  But there is
> nothing to stop you from putting in a sync card (and driver) and run PPP
> over eg. a T1 talkig to a Cisco or something...

I would say that if you're running PPP over a T1, then your CPU
probably can't afford to be doing the compression, even if you wanted
it (which you probably wouldn't).

Something along the lines of:

When you have enough bandwidth to cause a problem for a user-land
implementation, you have enough bandwidth that you either do not need,
or cannot do, compression...

I'm not sure that that's true, but i don't think it's too far off.


> Wouldn't it be better to have the compression code optional with some
> compile time option to decide on wether or not you should have it in
> the kernel in the first place?

Certainly, that's much better than what's there now, i'd say.  I'm not
convinced that it should _ever_ be in the kernel, but i'd be willing
to say "this is good enough," because it at least it gives people who
can't afford the wasted space a back-off position.



cgd