[10971] in www-talk@info.cern.ch
Re: Compressing HTML
daemon@ATHENA.MIT.EDU (Henrik Frystyk Nielsen)
Tue Sep 9 22:30:56 1997
Date: Tue, 09 Sep 1997 22:01:48 -0400
To: Andrew Daviel <advax@triumf.ca>, www-talk@w3.org
From: Henrik Frystyk Nielsen <frystyk@w3.org>
In-Reply-To: <Pine.LNX.3.95.970909001918.4632B-100000@andrew.triumf.ca>
Resent-From: www-talk@w3.org
At 12:28 AM 09/09/97 -0700, Andrew Daviel wrote:
>I just wondered what if anything had happened to the idea of compressing
>HTML. On my Unix server with Unix Netscape I can serve a regular 123kb
>file, or the same one with Content-encoding: x-gzip at 3268 bytes.
>On Windows it doesn't work.
>I believe that gzip is required for compliant VRML viewers, and there
>is certainly a gzip binary for Win.x.
>Although many last-mile links like 28.8 modems use compression
>I believe that most backbone links, proxy cache etc. do not, so
>it would seem to be a useful feature.
Compression of HTML can be a big win in many situations. It can also be
improved if you canonicalize the HTML in certain ways before compressing
it. You can have a look at the performance work that we have done [1] at
W3C [2] and Jeff Mogul at DEC/WRL [3].
Libwww [4] has an implementation of on-the-fly decompression using the Zlib
library, which we used to make our measurements. ZLib is written in C and
is available for most platforms including windows.
Henrik
[1] http://www.w3.org/Protocols/HTTP/Performance/
[2] http://www.w3.org
[3] http://www.acm.org/sigcomm/sigcomm97/papers/p156.ps
[4] http://www.w3.org/Library/
--
Henrik Frystyk Nielsen,
http://www.w3.org/People/Frystyk
PGP:0x1F71508D
