This is mostly going to be about how and why things are not being
agreed. I was not physically at the meeting but was watching
and listening and occasionally interrupting over the MBONE.
Comments in [square brackets] are mine.
-------------------------------------------------------------
Eric Huizer started (as IESG member and one of the 3 Area Directors
related to the URI work) by giving suggetions. These were basically
a. Very important spec for future: should not make sacrifices for
current installed base.
b. URL document is almost not quite at consensus.
c. URN is not but should not hold up URL
d. Consensus is not achieved with a vote, chairs must get it somehow,
on the net, not necessarily in the meeting room.
e. Chairs should prevent newcomer questions from slowing progress.
f. IESG needs a Functional Specification to explain what the
URL spec should do so that they have something to hold it up
against. Without that
Eric as AD would stop it consensus or no.
[I wondered whether (e) could be applied to (f) ;-)) ]
Alan Emtage explained that a bar-bof of 5-6 people had a consistent
view of what Eric's document should say, viz:-
URLs 1. Pointers for dereferencing
2. Transient
3. machine consumable
4. transport-friendly
URLs Location independent
Persistent
Human Transcribable
Transport friendly
Machine consumable
URCs Identification, metainfo.
HUMAN consumable and also machine consumable.
Transport friendly
URL & URN caching.
[if 4 points is a functional spec]
John Kunze was a member of the group, which was only trying to
figure out how to make progress. The group introduced a new field.
[seriously, they did.]. This was a URL spec version number
to be quoted in each URL, like
url:1:ftp://info.cern.ch/pub
The group needed now to "define the entire UR* area,
including describing resolution services which they would come
up with." [This worried me]
Marshal Rose as IESG member wondered who string up, with a WG
taking 18 months on a 12 page document. He has something
funny on the backof his Tshirt which wasn't multicast.
Larry Masinter pointed out that 1. above wasn't a god definition, as
you can't really be rdereference a telnet: or mailto: URL.
[Agree. Better to talk about object model, with retrieval
not being the only possible action on an object]
There was a lot of discussion of terms. There was a lot of
voting by making noise, some of which I attempted to
join in with over the net.
Steve Putz [intelligently] pointed out that URLs could be
defined as those name spaces which are defined by a direct
mappoing onto protocols. Cliff Lynch confusingly
suggested that URLs should be as indendent of protocol
as possible. [I hope he meant URN in which case I agree]
Dave Crocker asked that the group consider three real problem
scenarios. [This to me emphasises that IETF ancients and
newcomers were in the same boat in that they needed the scene
setting and the foundations of what wee are talking about defined.]
Larry compared URLs to MIME external body parts and noted
differences.
At this point it became more difficult to take notes and I stopped.
By the end of the meeting, Jim Fulton had volunteered to write
a requirements spec for the URL document.
[Now I am going to put down some of this functional spec myself.
Watch this space.]
Tim