9 December 2005

Attendees
---------------
Frank Siebenlist
Hiro Kishimoto
Andrew Grimshaw
Marvin Theimer
Mark Morgan (Notes)
Dave Berry
Manuel Pereira
Dave Snelling
Mike Behrens

Agenda
-----------
* Review Assumptions in document
* Use of WS-Address To field
* Assumptions of what in headers on the wire
* Globally Unique AbstractNames and ramifications
* Frank's Email

* Assumptions that went into WS-Naming
	- Most tooling pretty much assumes that the To field is a URL, not a name
		-- not clear that this is going to change
		-- this is a critical assumption
	- Requirements
		-- Some kind of unique string to identify resource endpoints with a high probabily (for equality comparisons on EPRs)
		-- Desire to rebind an EPR for whatever reason
		-- Two seperable things
* WS-Addressing "To" field
	- Abstract Name as a piece of meta data would not flow over the wire.
	- It may not be there.  If the dispatch on the other end NEEDS that information, then it should encode that information somewhere else where it would be passed on the wire.
	- Same thing applies to the intermediary -- if you need information, then you need to encode that information elsewhere
	- Security policies
	- If you construct the unique name from To and ReferenceParameters, then you loose location transparency
	- What about EPRs that are regenerated exactly (with different AbstractNames)?
	- Marvin's compromise
		-- We should recommend/require that the minters of EPRs with AbstractNames deal with the re-use problem.
	- A profile may be necessary that says in order to avoid un-inteneded aliases, there be a proflie on address and reference parameters that allows for backends to distinguish between re-used addresses
* How do we "guarantee" globally uniqueue abstract names
	- Minters should use well known techniques for generating globally unique abstract names
	- There are ways of doing this
		- RFC4122
		- GUIDs with additional information
	- Do we want to provide guidance? or do we mandate?
		- recommend against mandating
	- Everyone agrees that unique in space and time is good
* Is the spec. OK, or do we need to add things to the spec.
	- Recommendation text
	- Inform users of abstract names that they must include all necessary information for dispatching into other places in the EPR
	- Write up a profile for how to use WS-Addressing "safely" (i.e., deal with un-intential duplicate address uses).
	- Action on the working group to write this profile or punt it into OGSA