W3C home > Mailing lists > Public > uri@w3.org > September 2005

Re: XMPP IRIs: feedback requested

From: Peter Saint-Andre <stpeter@jabber.org>
Date: Wed, 07 Sep 2005 15:07:07 -0600
Message-ID: <431F567B.4030305@jabber.org>
To: Joe Hildebrand <JHildebrand@jabber.com>
Cc: uri@w3.org, xmppwg@jabber.org
Joe Hildebrand wrote:
> 1)
> 3920 references 3454 (Stringprep), which references ISO10646 and Unicode3.2,
> not UCS2.  There's lots of stuff that's not in the BMP that are still valid
> characters in a JID.

Will correct.

> Other than that, I think the intro is very good.


> 2.2)
> I think there's a syntax bug in the ABNF.  hier-xmpp in the xmppiri
> production doesn't have a definition.  Do you mean ihierxmpp?

Yes, sorry.

> Should '%' be in nodeallow or resallow?  Probaby not, since that messes up
> pct-encoded.  If not, fix 2.7.2 and 2.8.2 as well.

They are allowed by RFC 3920 but shouldn't be allowed in XMPP URIs/IRIs 
for the reason you adduce.

> The resallow production makes me realize, how come "&'<> aren't disallowed
> in resourceprep?  Don't they have the same XML consequences as in node?

Sounds like a matter for rfc3920bis.

> Using section 3.2 of 3987 to convert to URIs means that IDNs get %-encoded,
> not punycoded.  My understanding was that this was Leslies original beef.
> How come she didn't have this objection to 3987?

I think her main concern was that draft-saintandre-xmpp-uri-* defined 
its own rules and didn't re-use what's in RFC 3987.

> 2.3)
> Can we have language that says that the auth section is optional if you
> already have default credentials?  I think that's right, but am willing to
> be talked out of it.

Yes, that sounds right.

> 2.5)
> Do we ever need to do more than one action at a time?  Example: presence-sub
> and roster update for nick.  I suppose we could update JEP-147 so that
> ?subscribe or ?roster or ?something-new does both. ?rostersub;name=Romeo for
> example.

Sure. I don't see a strong need to allow multiple actions at once -- 
that opens the can of worms a little more widely than I'd like.

> Can the examples use "example-node" rather than "random-node"?  They aren't
> really random, and I don't want people to think they can send to a random
> person this way.

Sure, that makes sense.

> 2.7.3)
> (nitpick) Maybe a note that points out that &#x159; and &#x10D; are already
> in stringprep canonical form?  Maybe it will save someone else from wanting
> to double-check... :)


> 5)
> Maybe mention SPIM due to harvesting (same as mailto)?
> No passwords in the iauthxmpp section, unlike http: or ftp:.  Aren't needed
> since we have SASL ANONYMOUS.

Good point, I'll add that.


Peter Saint-Andre
Jabber Software Foundation

Received on Wednesday, 7 September 2005 21:07:06 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:09 UTC