Registry Working Group

Kickoff Meeting: London: 19-20 March, 2003

Contents:

Registration

Please register at: WhosComingToRegistry19032003

Agenda

Date	Time
19th March 2003	10:00	COFFEE/TEA available
	11:00	Introduction	Tony Linde
	11:15	Scope - discussion	TL
	12:00	GLU	Andre Schaaff
	12:20	OAI	Andy Powell
	13:00	LUNCH
	14:00	JINI	Marco Leoni
	14:20	Metadata Requirements	Anita Richards
	14:40	'Must have' Requirements - discussion	TL
	15:30	COFFEE/TEA
	16:00	Coverage	Clive Page
	16:20	AstroVirtel	Alberto Micol
	16:40	AstroGrid Registry	Elizabeth Auden, Keith Noddle
	17:00	Work Packages - discussion	TL
	17:30	FINISH
20th March 2003	09:00	COFFEE/TEA available
	09:45	Introduction	TL
	10:00	WP1 - discussion	all
	10:30	WP2 - discussion	all
	11:00	WP3 - discussion	all
	11:30	WP4 - discussion	all
	12:00	WP5 - discussion	all
	12:30	Next Steps (May meeting)	TL
	13:00	FINISH

Note:
40 min presentations will include 15 mins for Q&A
20 min presentations will include 5 mins for Q&A

Location

London e-Science Centre, 19-20 March 2003
South Kensington Campus
Imperial College London
SW7 2AZ
United Kingdom

Travel info

• Map
• Imperial College hotel
• South Kensington hotels
• Bayswater hotels

Notes

The meeting kicked off on time but was delayed by problems in getting the telecon working. In the end, Bob set up a NASA telecon link which we used for the rest of the two days. Because of the delays, we fudged the above agenda, dropping all discussions and running the presentations one after the other.

19 March 2003: Presentations

A number of people made presentations on topics related to the VO Registry. I scribbled down a few notes during a couple of the presentations and have added them here. Maybe others who attended can add to these.

GLU

presented by Andre Schaaff: download presentation from below

If we apply the GLU paradigm to REGISTRIES:

The structure would be:

• there can be several DOMAINS
• there are a number of running registries (CLIENTS and/or SERVERS for one or several domains, each running a GLU-like deamon)
• each registry uses as a hook, or entry point to the system
• the different servers for a domain are listed/authenticated by a MANAGER

Each client has a local dictionary (list of items in its local registry) and also a local copy of all the other definitions in other registries for the domains he has subscribed to.

Update mechanism and propagation:

• when a server updates its local dictionary, a synchronization mechanism propagates the modification on all dedicated clients

In this scheme:

• Each data center can be in charge of maintaining locally the description of its own services in its local registry.
• Services could also be registered via a web interface, on a dedicated deamon, for people who don't want to run a local deamon.
• When a change is done locally, it is automatically propagated to other registries sharing the same domain.
• Each registry has a copy of other registries' definitions in the same domain.
• The system can work whatever the metadata adopted for the actual description of the registry are.

• questions about UDDI:
  v2 is better and is worth looking at
  there are not enough hooks in it for extensions

OAI

presented by Andy Powell: download presentation from below

• Andy was member of the Tech Committee which came up with OAI-PMH
• not comparable to GLU but might be used in conjunction with it
• its roots are in e-print archives
  • now a more generic standard
• goal was to harvest metadata into central services
• 10 key points (see slides)
• terminology: resource, item, record
• protocol describes 6 request types
• utilises Dublin Core (DC)
  • DC very flexible
• JIESR: JISC Information Environment Service Repository: more akin to VO Registry

In the following discussions, Andy mentioned the idea of using a harvesting method for portals to maintain a cache of registry information so that they do not have to interogate the Registry for every query.

JINI

presented by Marco Leoni: download presentation from below

• approach to 'self-healing' network
• lookup service akin to registry
• metadata can be associated as an object
• allows search templates
• issues are that it is Java-specific and is designed for local context

It was agreed that, whether the technology can be deployed as a Registry solution or not, the design may have features that we can adapt or adopt.

Metadata Requirements

presented by Anita Richards: download presentation from below

• archive catalogs can refer to tabular data, images, spectra
• key issue: how do we get metadata from the catalogues, headers etc into the registry
• need extension of UCDs to cover eg frequency resolution
  discussion here determined that the resolution should be a value associated with a UCD rather than each resolution having a separate UCD
• data quality is another issue: how is it determined and recorded
• UCDs can evolve from use cases and usage

Coverage

presented by Clive Page: download presentation from below

• even finest grained registry will only require 1.0E+07 entries (per observation per instrument)
• how to categorise wavelength: low/high?
• need to manage unit conversion
• sparse coverage: ranges, bitmaps, ...

AstroVirtel

presented by Alberto Micol: download presentation from below

• Science Case (De Grijs)
• developed work plan
• constraints based on distance, data quality (how to define quality)
• need creation of data model for registry

AstroGrid Registry

presented by Elizabeth Auden: download presentation from below

I was out of the room for most of this presentation: perhaps others can contribute. My notes relate to the following discussion.

• need a registry data model
• should be able to refer to other entities: non-services

Discussion

We began discussing 'Scope' of the Registry WG.

First thing agreed was what the Registry registered: services or resources. We went for Resources as a more generic name for the entries.

The scope discussion didn't get much further so we agreed to continue the next day with discussion of the seven work packages that I'd defined at the end of my first set of slides.

20 March 2003: Work Packages

The discussion of Day 2 concentrated on each work package, what it should cover and some initial goals. As it happened we dropped from the initial seven WPs to five, as follows. I tended to update the Day 2 slides as we went along so they reflect what was agreed for each WP.

Rwp01: Management, Policy & Documentation standards

Scope:

• produce documentation template
• act as liaison with IVOA exec
• coordinate WP plans and publish integrated set
• look after policy issues (eg how do Data Centres join the VO registry)

Rwp02: Requirements, Science Cases, Use Cases, Test Cases

Scope:

• 5-6 science scenarios
• use cases
• requirements

Discussion about testing focused on issue of how to 'validate' a registry implementation. Can it be done with a script of tests and expected results or can it be more specific, ie have set of test resources which are added to the registry and baseplate queries are expected to produce set results?

Rwp03: Metadata specs

Scope:

• define data model (structure and content) for information entered into registry
  • need accompanying explanation
• define xml schemas for minimum but extensible information in registry (like DC)

Rwp04: Registry replication and interfaces

Scope:

• security/privacy of metadata
• how is metadata harvested (into registry & from registry into eg portal cache)
• query schema
• maintenance (add/change/delete) schema

There was some discussion about how data centres 'join' the VO registry (or set of registries); this prompted the extension of Rwp01 to include policy issues.

Rwp05: Implementation coordination

This WP began life as one to create a relatively simple reference implementation. It was felt that this was unnecessary given the number of projects underway.

Scope:

• coordinate implementation of registry standards among VO projects
• ? create software library
• ? develop components

Conclusion

• all communications should be via the Registry mailing list
• any message on the list should be prefixed with the wp code, eg 'Rwp03: look at my data model'
• wp leads need to be assertive about getting others to work on topics
• all Scope documents and Plans to be produced by 11-April
• TL to update wiki with notes from this meeting (DONE: finally)

Attachments

see below