RE: HTTP Compression & Squid

From: Jordan Mendelson <jordy@dont-contact.us>
Date: Fri, 18 Dec 1998 14:19:16 -0500

> The cache is really the wrong place to use gzip. gzip is optimised
> for high compression at the expense of compression time (bzip2 even
> more so); therefore it is best used in compress once, uncompress many
> environments, i.e. compress on the source server, before storing, and
> conditionally uncompress in transit. (Uncompressing is deterministic
> and linear on uncompressed size.)

So why not simply compress all data which goes into the cache and like you
said, uncompress it for clients which do not pass the Allow-Encoding or
Transfer-Encoding headers.

This would be far better than compressing an object over and over and over
again, since gunzip uses far less CPU than gzip.

The side benefit is that you could store a lot more in your cache :)

> LZ77, I believe, predates LZW (compress) and I suspect LZW was
> invented for this sort of on the fly compression even though LZ77 is
> better.

Too bad Mozilla only supports gzip so far :)

Jordan

--
Jordan Mendelson     : http://jordy.wserv.com
Web Services, Inc.   : http://www.wserv.com
Received on Fri Dec 18 1998 - 12:14:40 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:41 MST