summary of recent RI discussion

[Paul Harrison]

Tony Linde wrote:
> Sorry, Paul, I'm getting confused about your 'searchable'
> registry/authority. A full registry is one which supports the query
> interface and which contains *all* resource records from *every* authority:
> every full registry is identical in the records it contains (subject to
> harvesting overlaps).

That is the premise that I am saying will not be 100% true in realizable 
deployments - I do not think that we will be able to achieve this unless 
we formalize the storage format for the registries.

> 
> Saying that a registry contains 'searchable' authIDs does not make sense as
> every searchable (full) registry contains *all* authority records and all
> resource records.

what is being proposed here is not incompatible with this vision though 
- the searchable authorities list would just
> 
> 
>>A full registry might have the entries
>>
>><Resource xsi:type="Authority">
>>   <identifier>my.authority</identifier>
>>   <OwnedByID>ivo://my.authority/Registry</OwnedByID>
>></Resource>
>>...
>><Resource xsi:type="RegistryType">
>>   <identifier>another.authority/FullRegistry</identifier>
>>   <searchableAuthority>my.authority</searchableAuthority>
>>   <searchableAuthority>another.authority</searchableAuthority>
>></Resource>
> 
> 
> Firstly, I still think we need both ownedBy and managedBy authIDs to
> differentiate between those IDs which a registry can and cannot change (can
> change owned resources, cannot change managed resources).

I do not understand what "managed" means in this context. If you mean 
that a publishing registry supplies records to a full registry, which is 
  then free to alter the records, then that is a publishing model that 
would have horrendous synchronization problems. The whole point of 
publishing would be to have a single source of the information, so the 
publisher would update their registry (or whatever was implementing the 
harvestable interface) whenever they had new information, and expect it 
to be propagated via harvesting - they would not welcome having to 
manually resolve merge conflicts with the full registry.
> 
> Secondly, I don't think that publishing registries ought to have registry
> entries. They are not VObs resources since they are not accessible to the
> VObs. When a publishing registry links up with a full registry it is a
> private matter between them - no-one should ever know about the publishing
> registry. It might be using registry software to make it easier for its
> records to be uploaded into the full registry but it is not really a
> registry: it is not a VObs resource and should not be listed as such.

if it is not really a registry that was supplying the data, then it 
would not be listed as a registry - the "full" registry would then be 
listed as "owning" the authority id.
> 
> 
>>The reality is that not all registries can 'support' 
>>arbitrary extensions even in this limited extent - even the 
> 

-- 
Paul Harrison
ESO Garching
www.eso.org