W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > May to August 1996

Re: Unregistered charset values in HTTP 1.1, the ISO-8859-* values

From: Roy T. Fielding <fielding@liege.ICS.UCI.EDU>
Date: Wed, 10 Jul 1996 19:26:13 -0700
To: Olle Jarnefors <ojarnef@admin.kth.se>
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, iesg@ietf.org
Message-Id: <9607101926.aa16210@paris.ics.uci.edu>
> I wouldn't go that far. Certainly for experimental
> purposes or within a certain community a character set
> that still hasn't been registered should be possible to
> use. But in that case the use of a charset value starting
> with "x-" (and therefore guaranteed to never be
> registered by IANA) should be required by the HTTP 1.1
> specification, as it always has been in MIME
> specifications.

No. The MIME specifications have always been broken for that reason.
RFC 1521 required behavior that no implementation ever obeys --
that being a check of the IANA registry before transmission of a token 
which is not prefixed by "x-".  Requirements which serve no useful purpose,
and which are directly contradicted by all known implementations, do not
belong in any RFC.

Even if such a requirement were added, it is poor engineering to implement
a system that relies on "flag days" in which to switch from existing use
of an x-token to use of its registered non-x equivalent.  The result is that
a good implementation treats any "x-" prefix as if it didn't exist, so that
the same application won't die when somebody eventually gets around to
registering that particular token.  In other words, a good implementation
does not distinguish between registered and non-registered tokens, and at
that point there is no reason to use the "x-" prefix for anything.

Our experience with "x-" prefixes has been uniformly bad; rather than
encouraging registration of all tokens, they encourage the easy route
of just using the x-token as the standard for that type.  I see no reason
to further encourage their use.  IANA should certainly continue its
normal behavior of never registering "x-" prefixed tokens, but that
registration process is outside the scope of HTTP.


 ...Roy T. Fielding
    Department of Information & Computer Science    (fielding@ics.uci.edu)
    University of California, Irvine, CA 92697-3425    fax:+1(714)824-4056
    http://www.ics.uci.edu/~fielding/
Received on Wednesday, 10 July 1996 19:33:03 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:05 EDT