- From: <Harald.T.Alvestrand@uninett.no>
- Date: Thu, 14 Sep 1995 09:13:21 +0200
- To: Roy Fielding <fielding@beach.w3.org>
- Cc: Larry Masinter <masinter@parc.xerox.com>, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Roy,
I agree with you to a large extent, but disagree in the end result.
- User agent capabilities need to be described in a non-system-dependent
manner.
Agreement.
- I think that unless we can show that this description can usefully be made
the basis for (or derived from) a configuration file in at least one common
environment, we don't have an useful description technique.
So, I think that Nathaniel's approach of standardizing the labelling
(using MIME) and publishing an informational on the configuration file
(mailcap) was just about right. The problem now is that the configuration
file documented is inadequate for handling the labelling, so we don't have
"running code" proving that the features of the labelling scheme can be
used in practice. I think we need that.
(See the ad nauseam discussions about whether it is sensible to define
application/multi.level.subtype as long as mailcap doesn't support
application/multi.*, or the problems with configuring a multipart/signed,
for examples of what I mean about inadequacy)
harald A
Received on Thursday, 14 September 1995 00:19:40 UTC