Content transformation by proxies

First, set me state that I can live with the treatment of content
transformation in the 05 HTTP/1.1 draft, in particular because there
is a warning header to mark transformations, which makes it possible
for me to detect and avoid any proxy transforming content.

The comments below should be taken is input to a more general
discussion on the desirability of transformations.

jg@w3.org:
>
>I'd have more sympathy to saying transform-allowed as opposed to
no-transform
>if it weren't for the following two observations:
>
>1) the default case should be the kind case to the net; transforms can help
>on bandwidth usage.

Transforms which downgrade image quality to reduce bandwidth are _not_
by default kind to the net.  They are *evil* to the net.  One of the
fundamental advantages of a digital medium like the internet is that
long-distance transport does not affect the quality of the content.
Default bandwith-reducing transformations would eliminate this
advantage.

I regularly work on a sun with a black/white (not greyscale!) screen,
and found that it is impossible to use this machine to browse most web
sites that use clickable map and button .gifs.  When these .gifs are
dithered (transformed) to black and white, you can't read the text in
them anymore.

Many site designers work hard to make their button and map .gif files
as small as possible, but no smaller.  A proxy trying to reduce
bandwidth by transforming all of these .gifs to smaller .gifs (or
.jpegs) which are 50% smaller will simply render the site unreadable.
If such proxies become common, site designers will be forced to start
making their .gif file sizes twice as large, and possibly larger, to
ensure a readable result for the general user.

>  Those that have legal/health/scientific reasons to 
>forbid it can do so, but most won't have to worry about it.  Most of the
>time it is a good thing.

Most of the time it is a bad thing. See above.  Find a box with a
black/white screen and start browsing the web if you don't believe me.

>2) you can't presume that data won't be transformed today;

I do presume that my data won't be transformed, and I want to keep
presuming it.  Why good is a digital medium if I can't even assume
lossless transmission?

As for the AOL case, which is the only example of lossy
transformations I heard of: I never assumed that AOL actually
implements access to the WWW part of the internet for its users.  AOL
seems to implement some approximation of web access that is (or at
least was) deficient in showing .gif pictures and that sometimes shows
old versions of un-cachable things.  When I, as a European, am making
web content, I don't care about breakage introduced by AOL: precisely
because it introduced such breakage, I define AOL as outside of the
internet I am trying to serve.

As a first step to making content transformation by proxies ready for
prime time, we would need a broad discussion about acceptable
transformations on www-talk.  Due to time constraints, I won't start
such a discussion right now, however.

>                                - Jim

Koen.

Received on Saturday, 8 June 1996 08:31:29 UTC