I consider out SIGCOMM paper the "best possible" benefit from HTTP/1.1,
rather than the "typical"; but also note that changing the site by factors
of two or four the other way would not have made a huge impact on our
results, either, as far as I can tell. Since we had no data on what
"typical" actually was at the time, the best we could do was generate
something toward one side (showing the best it might do, rather than some
typical that might not be typical).
Also note, however, that we also ignore what "click ahead" to additional
pages on the same web site might bring; that is a benefit of 1.1 that
is pretty common that we ignored. (Hopefully real traces will catch
alot of that).
We hope to be using real traces on our further HTTP/NG work; our
only future use of the Microscape site will be to intercompare limiting
behavior of HTTP/NG with HTTP/1.1, rather than use as a "typical" web site
(which we never thought the Microscape site was...).
So for Gawd's sake, don't use the Microscape site as "typical"; it isn't.
- Jim Gettys