Some further replies in-line.
On Fri, Jan 21, 2011 at 9:37 AM, Julian Reschke
On 21.01.2011 17:57, Ted Hardie wrote:
Reminder: the reason this was written down was so that
can be specified as XML system identifier in HTML5
This rationale isn't in the draft, nor is the token legacy-compat.
...because HTML5 defines it... (I think)
But the question with this how you will get interoperability. If there is a
token registry, then these should populate that registry pretty much
as soon as the registry opens, as these are entries for which conflict
has a bad result. The whole point of a registry is to avoid collisions in
the token portion of the identifier space.
at the section you reference, I see it also defines about:srcdoc as
unresolvable URI. It should be included in this doc, if it goes forward.
I don't think the goal of the registration document is to define all about
URIs out there. (and I don't think it should).
It goes to significant lengths to define the behavior of about:blank,
and it mentions but does not define others. That's a bit odd, in
IETF circles, especially when you are defining a protocol to have
"reserved" and "unreserved" tokens. This iis why the suggestion
of a registry is being made. That gives a place to look
for behavior definitions and confirm reserved vs. unreserved.
If that is not really wanted, this draft is reserving a URI scheme on
what looks to me to be closer to provisional terms.
I have enough experience on the URI registration side to understand
that people mint schemes without registration pretty readily, and minting
tokens seems even easier. If you don't care at all about interoperability
in this behavior, a registry is not needed, but then I'm not sure why
a permanent URI registration is needed either. The about: convention
has been around a while, and a provisional to note it seems enough.
That said, I note that HTML5 has a number of what it calls "willful
of the URI spec, in which it counsels the reading who actually knows what
the spec says to pretend it was using a term other than URI. (One of
is just past the fragment identifier used above). Most uses of about
are outliers in the URI world by a long chalk. Why not simply define
as a different identifier form that happens to have a colon in it
it is) and make its use as system identifier in HTML5 a "willful
the XML spec? That seems entirely consistent with the document's modus
and save IANA the trouble of setting up a registry.
Browsers use "about:" where URIs are entered. So no matter what we think, it
will be very hard to ever register a "about" URI scheme for something else.
I agree, and a provisional to block its registration seems sensible. But if you
are not eventually trying to get interoperable behavior, having "reserved"
elements for this scheme seems, at best, confusing.
Why not just accept it, and have a registration that at least enhances the
situation (as opposed to it not being registered)?
The last call invites advice from the community on how to improve the document.
Why not just accept, and trust it enhances the situation?
Also -- the spec doesn't define an IANA registry, so there won't be any
trouble for them anyway.
Best regards, Julian
Ietf mailing list