What URI for BCP 47?

[Frank Ellermann]

Richard Ishida wrote:

> I still have no clear idea which is best.  Help :-)

Working with BCP numbers is not yet the usual case,
but you obviously need it while waiting for 3066bis
or its successor.

 From the many places where you can get RFCs only the
RFC editor site is "canonical", and it offers at
least three access variants:  FTP text, HTTP text,
and its search engine (human readable meta data plus
links).

A BCP is necessarily "IESG", therefore the IETF site
is also "canonical".  IIRC you asked for a HTTP URI,
otherwise we could check out the IETF URN, maybe it
offers a way to specify BCPs (?)

> ftp://ftp.isi.edu/in-notes/bcp/bcp47.txt

IMO that can't be canonical, you're not interested in
questions who acts as actual RFC editor, and how the
current RFC editor and the IETF decide this.

> ftp://ftp.rfc-editor.org/in-notes/rfc3066.txt

That's the RFC, as soon as 3066bis got its number you
wouldn't want its predecessor.  Addditional problem:

At the moment the published BCP 47 is _one_ document
RFC 3066, but the not yet published approved BCP 47
consists of two RFCs, 3066bis and matching.

> http://www.ietf.org/rfc/bcp/bcp47.txt

That's where it hits you, in theory.  In practice you
will probably get what you want, 30066bis or better,
not the matching part.  "They" have been warned many
times that more than one RFC per BCP might be a bad
idea, but "they" insisted on it (because the matching
stuff replaces a single sentence in the old RFC 3066).

> http://www.rfc-editor.org/rfc/bcp/bcp47.txt

Same problem, otherwise the other "canonical" source.

> http://tools.ietf.org/html/bcp47

Same problem.  Otherwise that's what I'd use, behind a
PURL maybe, as protection against future changes.  It's
apparently a volunteer site controlled by the Chair of
the IETF "tools" team hosted by his employer, in other
words nobody promised that it will exist "forever" (as
long as the IETF exists).

> Who at the IETF should I ask this question of?

Brian, the Chair.  I doubt that he has better answers,
but this multi-RFC-per-BCP concept was his idea (not
for BCP 47, he needed it for another BCP), and he is
also in the position to judge the persistency of the
tools server.

IIRC the W3C uses some PURLs, "roll your own" could be
a solution.  But asking Brian if he has better ideas
can't be wrong, please forward his reply if possible.

Frank