Scope of registry

[Marco Leoni]

Clive Page wrote:

>On Thu, 6 Feb 2003, Roy Williams wrote:
>
>>In the OAI model, there are exactly six verbs that the registry understands.
>>The verbs are:
>>
>>Identify (who are you)
>>List Metadata Formats (what schemas do you have for metadata)
>>List Sets (collections hosted by this repository)
>>List Identifiers (return just the identifiers and dates of most recent
>>change)
>>List Records (return full records)
>>Get Record (return a full record from its identifier).
>>
>>This is a pretty basic set of services!
>>
>
>That seems to me rather more basic than will be suitable for what we want.
>I confess to thinking of the registry in database terms.  Users might want
>to send queries to the registry equivalent to this sort of thing
>(expressed for convenience in pseudo-SQL):
>
>SELECT * FROM registry WHERE WAVELENGTH BETWEEN "400 nm" AND "800 nm" AND
>  POSITION("12:34:56,-01:23,10") AND TYPE = "photometry"
>
>If we don't allow such selectivity in registry queries, won't all queries
>just return all records?  I thought the idea of the registry was not just
>to have a definitive list of resources, but also avoid the necessity for
>users to send their queries to _all_ registered resources, just on the
>off-chance that each of them might have some information of relevance.
>
In fact the registry should do most of the work for you, if we think of 
it as Tom suggested.
The registry should use the cached information to find out which service 
may have the answer then it will send the detailed query to that/those 
service/s.

>
>
>Querying all resources will be very time-consuming, and will also have to
>be repeated, because not all resources will be on-line at all times.
>Users who suspect that the archives of the University of Erewhon
>Observatory has something relevant, but find that it is frequently
>off-line will have to try it many times.  If, on the other hand, their
>initial query to the Registry shows that the answer concerning the Erewhon
>site is "no" rather than "maybe", they can avoid such time-wasting
>activities.
>
Perhaps asynchronous calls will help to reduce disocomfort due to 
unavailables services.
But this is related only to the second step: if the registry knows from 
the cached data that service doesn't help then the answer is immediately 
"no", otherwise you have to wait till the service will be online again.

>
>
>To those who doubt that this is an important factor I would suggest: go to
>the Astrobrowse site, http://heasarc.gsfc.nasa.gov/ab/, and select a full
>search, and see how long it takes to get an answer from all the linked
>resources.  I find it takes typically 40 to 60 seconds, but that sometimes
>one or two sites never respond (we all have occasional outages, after
>all).  Astrobrowse currently only queries the principal sites, all
>well-networked and well-managed.  When our Registry includes sites less
>well endowed, things may not be as good.
>


Cheers,
    Marco
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ivoa.net/pipermail/registry/attachments/20030207/6938acee/attachment-0001.html>