I think we're roughly on the same track regarding 1-5; I'll work on
straw-man spec text in this direction.
Regarding the registry itself -- see previous discussion about
relative merits and difficulty.
Regarding the licensing of the registry -- I can't help feel that this
is a red herring, given that IANA also manages the DNS root zone, IP
address and AS number allocation, and just about every protocol
registry associated with the Internet, including media types, Ethernet
numbers, HTTP status codes, IMAP capabilities, IPv4 and IPv6
parameters, language tags, port numbers, protocol numbers, PGP
attributes, SSH parameters, TLS parameters, SNMP number spaces, syslog
parameters, URNs, URI schemes, etc. See: <http://www.iana.org/
protocols/> for the full list.
In short, pretty much any "free software" that connects to the
Internet is riddled with dependencies upon IANA registries. On any
Unix-like system, take a look at /etc/services and /etc/protocols, for
starters. If this is a real issue, it affects much more than the
registration of link relation types, and you should take it up with
IANA, the IESG and the IAB immediately, or change your concept of
what's necessary for "free software".
Cheers,
On 11/12/2008, at 5:11 AM, Henri Sivonen wrote:
On Dec 10, 2008, at 04:56, Mark Nottingham wrote:
Thoughts (everyone, not just Anne)?
I suggest what I suggested at the HTML WG f2f at TPAC this year:
1) Rel keywords are are strings that are one or more Unicode
characters in length and don't contain HTML5 space characters and
don't start with the string "http://www.iana.org/assignments/
relation/".
- It follows that both "foo" and "http://example.com/ns/2008/12/
bar" are possible keywords.
2) If a keyword contains a colon, the keyword must (for document
conformance) be a valid IRI.
3) Keywords are compared ASCII-case-insensitively, that is, a-z
equals A-Z and otherwise code-point-for-code-point.
4) The Link HTTP header can use those rel keywords that are ASCII-
only (since HTTP header values aren't robust with arbitrary Unicode).
5) Consumers who want to see the world as IRIs (not browsers but
e.g. GRDDL extractors) run the following algorithm to convert a rel
keyword to an IRI:
a) Replace A-Z with a-z.
b) If the result does not contain a colon, prepend "http://www.iana.org/assignments/relation/
"
6) Keywords that don't contain a colon must be registered on the
WHATWG wiki (or a W3C wiki if the W3C decides to host the registry).
7) The registry wiki is licensed under a GPL-compatible but not in
itself reciprocal Free Software license (e.g. MIT or New BSD).
(Notes about points #6 and #7: I consider the IANA registration
procedure too cumbersome compared to editing a wiki, and I think it
is a bug that IANA registries are not under a Free Software license,
since not being under a Free Software license is a problem for
inclusion in Free Software.)
--
Henri Sivonen
[EMAIL PROTECTED]
http://hsivonen.iki.fi/
--
Mark Nottingham http://www.mnot.net/
