[e2e] Compression of web pages

[Vernon Schryver]

> From: "Woojune Kim" <wkim at airvananet.com>

> ...
> > http://online.wsj.com/article/0,,SB1029974251705867835,00.html ?

> ...
> From Mr. Mossberg's comments that the figures were grainy, I
> had the impression that some sort of proxy was taking the .gif,
> .jpg files apart and reducing their clarity / size , so as to make
> their sizes smaller. Wouldn't that work ?

I had not noticed those words:

] Sprint has done a great job of speeding up and optimizing the
] Internet connection so it feels faster than it really is. The actual
] link averages only 50-70 kilobits per second, far slower than a
] cable modem or DSL line, though a bit faster than a phone modem.
] But Sprint's network compresses all the data being downloaded, so
] it moves like a much faster system. I ran 21 tests of the download
] speed at four bandwidth test sites on the Web. The network performed
] like a system capable of 400 kilobits per second.
] 
] The downside is that Web graphics come out grainier and fuzzier than
] they do with a wired connection

That's an interesting thought.  Can unpacking a .gif, transforming
it, and then sending it as if it were the original violate copyrights
or trademarks?  Then there are the myriad other end-to-end objections
to hidden middle boxes.

> ...
> Overall no big difference in opinion. But Mr. Mossberg mentions
> that he used some of the bandwidth measuring websites. Do you know
> what sort of data do they use ? Would their data be more susceptible
> to compression ?

I don't know about you, but when one of those bandwidth measuring
sites reports a number for paths I know about that is not obviously
wrong, I'm somewhat surprised.  Some of them seem to be fairly accurate,
but others seem to exist to sell stuff.

If Sprint is rewriting pictures as they fly by, and if Mr. Mossberg
chose bandwidth measuring sites that use pictures to determine link
speed, then one would expect the their results to be match.

I wonder if an evil middle box could truncate most JPGs and GIFs,
say with a TCP RST, and get "grainy compression."


Vernon Schryver    vjs at rhyolite.com