Re: More notes on content negotiation

Dave Kristol (dmk@allegra.att.com)
Thu, 16 Nov 95 09:24:51 EST 

koen@win.tue.nl (Koen Holtman) wrote:
  [a long proposal on content negotiation]
  [...]
  > User-agent header based negotiation
  > -----------------------------------
  > 
  > One of the requirements I have for HTTP/1.1 content negotiation is
  > that variant selection based on the contents of the User-Agent header
  > sent by the user agent is efficient.  In my opinion, the v10-spec-01
  > negotiation text does not satisfy this requirement.
  [...]
  > - The mechanism should allow negotiation based on the User-Agent
  > header to be efficient.
  [...]

<purist> I really despise content negotiation based on User-Agent. </purist>

One problem with such negotiation is it precludes a user's
enabling/disabling features on the User-Agent.  Suppose someone decided
(Jeez, I don't know why) to disable tables in tables.  If my User-Agent
field says Browser/Version-with-tables-in-tables, this negotiation very
likely may decide to send a tables-in-tables version of a resource.
What then?

OTOH, I sat in a meeting this week with a content vendor that uses
User-Agent precisely to choose which content to send to a browser.  Not
just because of incompatibilities, but because there are users with
older versions of once-hot browsers that now don't have the latest
features.  The content vendor wants to send the best possible content
for each.  Makes sense from that point of view.

So, while I dislike the idea of some kind of feature profile, be it a
bitmap or whatever, that would seem preferable to just going on
User-Agent.

Dave Kristol