We are gzipping our html pages on hotornot.com now.. We've seen a
compression ratio of about 3:1. For browsers that don't support gzipped
content, we just send them the content uncompressed.
Haven't received any complaints from any users yet, although it has of
course increased the load on our machine. Since we ONLY serve html, and
don't serve any pre-compressed data like images, it is definitely worth the
trouble and the extra load. I'm not sure this would be true for most
websites, where images usually account for a high percentage of their
--- XMethods web service listings - http://www.xmethods.net
----- Original Message ----- From: <email@example.com> To: "Andy Armstrong" <firstname.lastname@example.org> Cc: <email@example.com>; <Morton@dennisinter.com>; <firstname.lastname@example.org>; <FoRK@kragen.dnaco.net> Sent: Saturday, March 24, 2001 6:09 PM Subject: Re: HTTP serving compressed content (was Re: Perl "competitor" Curl raises $52M)
> Andy Armstrong <email@example.com> writes: > > Gordon Mohr wrote: > > [snip] > > > That suggests to me that either (1) it's so transparent it's easy > > > to miss; or (2) it doesn't save enough in bandwidth costs to be > > > worth the trouble. > > > > It's generally completely transparent and pretty effective. In our tests > > we've seen typical compression ratios of 6:1 on typical HTML content. > > Last time I checked something over 90% of browsers hitting our sites > > were able to handle GZIP compressed content, and those that can't > > degrade gracefully. > > How do they degrade gracefully? > > FWIW, I found recently that <script src="foo.js.gz"></script> doesn't > work in Netscape 4.76; despite correct content-type and > content-transfer-encoding headers, the stupid browser tries to execute > the script before decompressing it. I don't recall whether I tested > this with MSIE. > > It seems libungif-generated GIFs could benefit from gzip > transfer-coding --- they'd be smaller than the corresponding LZWed > GIFs, and unlike PNGs, could be animated.
This archive was generated by hypermail 2b29 : Fri Apr 27 2001 - 23:14:52 PDT