- From: Larry Masinter <masinter@parc.xerox.com>
- Date: Mon, 29 Jan 1996 11:50:02 PST
- To: john@math.nwu.edu
- Cc: www-talk@w3.org
> Thanks for the fine discussion on just how hard it is to do content
> negotiation. An issue you didn't mention is the interaction of content
> negotiation and proxy-caches. This is one of the thorniest problems.
Actually, I think this particular issue is in hand, mainly because
'cache but verify' and 'don't cache' are available as fallback.
> Despite all its faults, content negotiation by user-agent seems to be
> emerging as the choice of major content providers. It is the only
> thing available today which meets their needs. No alternative is
> apparent on the horizon. I suspect we may have no choice but to live
> with it.
I don't accept that merely because this is the choice content
providers have made means that it is the best preference, or even that
"No alternative is apparent on the horizon". In fact, there are many
alternatives apparent on the horizon. They're just not 'in hand'.
That's what the work is all about: how to get closer than 'on the
horizon'. I believe all of the alternatives proposed have merit and
that we'll wind up with a combination of some (but probably not all)
of these:
a) documents with optional content
a1) marked sections & optional client side includes
(to be specified)
a2) embedded media marked with alternate contexts
(as appears in INSERT or EMBED)
b) use of new media type registration to describe media
with proprietary extensions that are not dealt with by
non-extended clients
(text/x-html-netscape-2.0)
c) content negotiation based on media type parameters
(extension of 'accept-charset')
d) content negotiation based on media characteristics
(accept-color)
e) content negotiation based on DTD components using modular DTD
> Microsoft seems to be embarking on a direct attempt to sabotage
> content negotiation by user-agent.
Never attribute to malice that which can be explained by ignorance.
Received on Monday, 29 January 1996 14:50:33 UTC