>Er, no, time to beat Greg with my metrics research hat. Microscrape
>is a collection of images (obtained from the early Microsoft and
>Netscape home pages) on a single page, thus making it twice the
>number of co-located inline images than two of the worst such pages
>on the Web. As nice as it may be to see positive papers on HTTP/1.1,
>I don't consider that to be a representative sample for normal
>Web content.
Whether it is any better or worse than other site is difficult to say, but
we don't pretend to say that the microscape site in any way shape or form
represent a typical Web scenario - you can see for your self at
http://www.w3.org/Protocols/HTTP/Performance/microscape/
This is why we now have the W3C Web Characterization Group to come up with
some real scenarios so that we have a much knowledge about what we are
designing for. This also helps prevents that we go off designing something
that a bunch of engineers think is a beautiful thing.
However, it doesn't change the fact that HTML compresses very well and that
because of the nonlinearity of TCP, compression of the first few packets
gives much more bang for the buck than further down the line.
One other thing is that the casing of HTML tags actually matters.
Algorithms like deflate favors lower case characters, for example. I did
some quick and dirty testing of this at
http://www.w3.org/Protocols/HTTP/Performance/#Compression
Henrik
-- Henrik Frystyk Nielsen, World Wide Web Consortium http://www.w3.org/People/Frystyk