2.1.18 Future of Uniform Resource Identifiers BOF (furi)

Current Meeting Report

Future of URI (FURI) BOF
IETF 50, Minneapolis
Tuesday, March 20, 2001.
Chair: Leslie Daigle
Official scribe: Eric Prud'hommeaux


Intro & Overview


- Coordination & publicization of existing work-
- Clarification of existing work
- URI services/vocabulary
- URIs and XML (e.g., as namespace names)


[Slide 2] Pre-History

- Long history of unregistered URI schemes and confusion about URL/URN/URI schemes and registration
- Sep 2000: XML Namespaces incompatible with relative URI references

[Slide 3] W3C URI Activity So Far

- July 2000: proposal for W3C URI Interest Group (feedback: "do a real Working Group")
- Oct 2000: W3C URI Planning Interest Group chartered to develop one or more WG charters.
- Teleconferences started in Jan 2001

I maintain a list of uri schemes that I've heard of.
Confusion from w3c membership: "should I invent a new scheme for X?".
Problems with trying to use relative URIs in XML references.

URI Planning Interest Group started Oct 2000; didn't actually get talking till janurary this year.

[Slide 4] W3C Context: Protocols and Services

- XML Protocol Working Group submission: SOAP

- Web Services Workshop, April San Jose

I'm here to do gap analysis, se the overlap and gaps between IETF and W3C.

[Slide 5] W3C Context: XML and RDF

- XML Core WG: maintains XML Namespaces chartered thru Sep 2001

- RDF Core WG: revising RDF 1.0 specs just starting

- XML-Core chartered to change namespace-spec if it ever changes.

[Slide 6] W3C Context: Clarification/Outreach

- Web Accessability Education and Outreach ongoing work in W3C since March 1998

- Semantic Web Activity launched Feb 2001

- Web Architecture Group in formation

In the name of accessibility, lots of folks are being taught to read documentation. I'd like to do that here.
We've noticed the cost of having no written web architecture.

Suggestions for Future Work

Graham Klyne: Clarification & coordination

1. Things that exist but need to be advertised and people directed at

I have no special expertise in URIs.
I've come to grips with the common misunderstandings.

Things that exist:
- The main resource is RFC2396. Re-reading this spec always reveals more.
- RFC2717 for registering URIs
- - There are many more, as Dan Connolly pointed out, that are not registered.
- RFC2141 for URN syntax
- RFC261 for URN namespace registration procedure

No plan to solve UR* controversies here: in particular, URNs (postion and location independent) any URI can be position and location independent. I think it's useful to have ones that say they are independent.

Many protocols use URIs at core.
- HTTP uses the URI bits to decide where to send the request.
- LDAP - uses the URI to decide where referral should go
- ENUM - return URIs for a given phone number.

Many document formats that use URIs
- HTML - used as essential linking mechanism
- XML namespace
- XPointer
- RDF - identify a description of meaning

2. Things that need to get fixed/resolved

Need to clarify and resolve historical misunderstandings:
- identification vs location
- naming vs addressing

Different understandings of the underpinning lead to different high-level conclusions.

This comes from my work in RDF, using URIs as very abstract things.

many forms of equivilence:
- ID
- Schema
- Resource
It is not always easy to say whether two uris are equivalent.

Roadmap of specs "comprehensive standard" Internet Draft.

Ways in which URIs might be used:
- how do you resolve a URI?
- how do you find out other information related to a URI?

Micheal Mealling: Issues in URI vocabularies & services

I'm interested in talking about existing work with URIs.
Based on exsiting internet apps, what new services do they need for reliable use?

Standardizing services that folks want to create around URIs

What does this URI identify?
- is this URI a cached alias for another URI?
- is this URI part of some collection?

These questions may be related to things W3C is working on but they are applicable to IETF protocols as well.

For instance, given an HTTP URI, I want to know if it's a streaming multimedia presentation so I can ask about quality of service.

What work needs to be done to make sure there's interoperability so the service as a whole is reliable?

DDDS (dynamic delegation and discovery service): Given a URI get some meta data about it.

ResCap: Who do you ask those questions, data format, low network impact (ResCap is over UDP, ie fast).

Dan Connolly: This is realated to sw, it's community stuff.

What can be done here?

There's some rough concensus about this goal:

- "Providing small chunk metadata to talk about the URI and its relationship to its Resource and other URIs/Resources"

That's what the interest group is proposeing we work on.

The next deliverable is a charter proposal.

There will be a proposed charter for standardizing these services coming from the interest group.

Question from the floor:
Why have a working group here as opposed to just W3C

Micheal Mealling: There is appliciability to apps the W3C is unconcerned with, LDAP, ENUM.

Is the intention to expand on the current URI definition?

Micheal Mealling: defining a new set of services, make sure all the services can use URIs.

Dan Connolly: Issues related to use in XML

[Reading from a displayed mail message]

Are these two bats the same bat?
I was surprised to see XML software that thought they were the same bat.

There have been long discusion of whether URIs are connected to reality.

I was hoping URLs would be useful in debugging stage.

Many folks want no obligation to serve these URIs.

After 17K mail messages, the conclusion was: "Doctor doctor, it hurts when I do this. Well don't do that."

W3C has hit the pause button - no new specs have relative URIs in namespace declarations.

Open Discussion

Leslie Daigle: This is now an open dicsussion.

George Michelson:
Worked with squid. Working with mirrorsoft.
Other WGs looking at abstraction of ...
employment of URI URN schemes. They have a low profile.
Some work in squid to look up URN.
Long way from context that people understand.
Papers use terms like friendly and short, only need persistent.
I'm in FUD space

Leslie Daigle: Yeah, the kinds of questions that came up over and
over were:
When do you develop a new scheme?
Are new schemes appropriate?
We need new clarification and publication of what's been done.

?3: I aggree whole hardedly.
... CDN (content delivery networks) ...

Mark Nottingham:
The charter talks about small chunks of metedata about resources.
In fact, large chunks are common, I found this in P3P.
Is this in scope?

?4: You would not put P3P in a service, instead have the service answer "Where would i get that P3P?"

Mark Nottingham:
At the time URNs were looked at in squid, registaiton had not been approved. There are documents that need to be read before saying URNs don't work. They work for me, I'm relaeasing a product in mid May.

Larry Masinter:
Please separate out the clarification work form the minting of services.

Micheal Mealling:
They are separate

Larry Masinter:
When we updated to 2396, we left behind material that needed significiant work -- description of several URI schemes, file: for instance. file: works differently everywhere. I would like help characterizing this.

2nd point:
There's been a draft for 4 years on i18n on URLs. In January, the name changed. We don't call them URIs. We are defining something new, IRIs (Internationalized Resource Identifiers). I would like to get this document out. [draft-masinter-uri-i18n-07.txt]

Leslie Daigle:
It is under consideration.
I'd like to take this opportunity to segue to the issue W3C coordination. We need to work with W3C. We need folks who have experience in these matters from the protocol aspect, too (IETF).

Dan Connolly:
IRI spec is descriptive of exsisting apps and many pending apps.
You need to respond within 3 to 18 months.

Dave Crocker:
I have been working in another WG working with UR<something>s.
The <something> options were apalling.
The community's ability (beyond that WG) to select <something>s and defend the choice is pretty weak.
We don't have clear definitions.

Leslie Daigle:
So what do you specifically want to ask the group?

Dave Crocker:
Do people use and hear others use these terms (UR{N,L,I}) with consistency?

Leslie Daigle does a hum poll:
Hums from those who say it's inconsistant have it overwhelmingly.

Jeff Hodges:
URLs are well understood in literature, but not URNs.

? :
If you ask if two docs are equiv, do you mean, is the identifier eqiv, or is the MD5sum equiv?

Micheal Mealling:
That will be explained.
That and "what is UR*" are the two biggest topics on these conferences.

Leslie Daigle:
We can't call for specific action items.
Who feels there is work to be done? .25 of the room
Who want to work on docs: .05 of the room

Larry Masinter:
Give names of the doc-workers to the scrib.

[grumblings all around, which lightens my load]



None received.