Re: Features

From: Gregory Maxwell <[email protected]>
Date: Mon, 24 Nov 1997 08:55:56 -0500 (EST)

On Mon, 24 Nov 1997, Cefiar wrote:

> There is a major problem here though when you are running a large cache and
> have run into the maximum your hardware allows, which I would definately
> believe is out there when you are referring to the Intel Architecture
> (mainly ram).

Yes, that would be quite a situation, you may wish to put up a second
cache at that point.. But software is cheaper then hardware..

But this is still more of a special case... As I said before, I dont think
most could use full compression, if you want it bad then make it.. I'm
sure some will need it.

> As for CPU hit, the only way I can see of truely avoiding this, is allowing
> some use of a plug in compression card. The problems here of course are
> that the few that are available most likely have very little support,
> and/or example code, and would have to vary for platform to platform, if
> some platforms have them at all!

Thats would prob be unnessassary.. Most caches are mostly cpu idle.. If
they use the LZO compressor they can alter the compression to fit the cpu
free time..

> However it gets developed, use some easy "replacable" way to implement the
> compression/decompression. Some type of negotiation header (at the start of
> a persistent connection) wouldn't go amiss either...

Yes.. This would be good.. It shouldn't just be for intercache. It should
also allow savvy browsers.. :)
Received on Mon Nov 24 1997 - 06:26:18 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:43 MST