roy at cacr.caltech.edu
Mon Feb 3 08:20:40 PST 2003
I am trying to push OAI. I believe it to be just the simple and
flexible standard that can federate our registries. Librarians use it
everywhere. The ADS uses it.
> (1) I strongly agree that we must keep the VO an enabling
> rather than an actual structure. Getting the right approach to the
> Registry is a key step in getting the philosophy right. We shouldn't
> over prescriptive. It should be possible to have multiple
> Registries should not require a supervising authority. A Registry
> be a service somebody can offer competitively, just like offering
> search services etc.
The OAI model is based on multiple registries (eg one at Edinburgh,
one at Caltech, etc). There could also be collector registries (eg
Astrogrid, AVO, NVO). It works because the protocol is *harvestable*.
You can ask for identifiers of all records that are more recent than a
given date. Therefore a robot can do this on a regular basis, grabbing
everything new from a group of other registries, so new records in one
registry percolate to others. A harvestable protocol means a
distributed virtual registry.
> (2) On the other hand, we want as much as possible of the software
> develop in the various VO projects to be inter-operable.
I think that the IVOA *must* agree on something harvestable, otherwise
nobody will be able to find what has been published.
The GLU system is made to be harvested. I think I know how to build a
OAI service on top of GLU. I believe that UDDI is also harvestable,
but I think the only things you can regiater in the registry are web
services. With OAI you can register anything you like.
> (3) Each VO project should be free to use its own choice of
> implementation technology, consistent with the interoperability
> principle, and also free to make design choices, such as whether
> registry is fine grained or coarse grained. (At this stage we
> want to positively encourage a diversity of choices, for technical
I agree entirely about implementation. I don't care if your brain is
made of cheese, so long as you say the right things! However, I point
out the large collection of open-source tools at the OAI site:
Once we have something harvestable, the next choices concern the
nature of the entities that are described in the registry, which then
crystallizes to the XML schema used to describe that entity. This
comes to the graininess issues. In a telecon last week, we (NVO)
thought it would be good to register a *table* (ie catalog) into the
registry, but it would not be good to register each *record* from that
table. (Librarians record books, not pages from books.) If you want to
record pages, then our robot will probably not harvest them. It will
only ask about the larger entities (tables).
In order to make a IVOA regstry of tables, we would need to agree on
the metadata description of a table.
> The conclusion is that the IVOA forum should aim at answering the
> question "what is the MINIMUM set of agreed standards that we should
> agree on" ? I think some of the US-VO resource data document is
> to this aim, and some goes beyond it. Likewise in AstroGrid we are
> starting to make some specific technology choices that we should not
> necessarily force on others.
What are these specific technology choices?
More information about the registry