ownedAuthority element

[KevinBenson]

Yes Matthew I am starting to have some other thoughts now.  But I do think
the hierarchical model would be nice, maybe make it flexible where we don't
have to use it.  Or similiar like you said possibly we decide to keep it
flat a little longer espcially if we are going to go through several
curation issues.

Here are some thoughts:
My thoughts are that Registries when harvested and need to return there OAI
form of there records will show <ownedAuthority> and there
<managedAuthority> records.

Hence much like today a Full Registry may not go the hierarchical route and
may just harvest the smaller Publishing registries if desired.  This might
also save us from some troubled areas of calling up registries telling them
to find a Full Registry to manage with.

Some registries may have <managedAuthority> and <ownedAuthority> of the same
authority id, in fact this is common in Full Registries.  A lot of our Full
Registries have there own authority ids that they publish.
Another good example might be CDS which is a publishing registry, I am
thinking now we don't require that a Full Registry manages "CDS" authority
id, CDS may wish to stay as a Publishing Registry and have
<managedAuthority> and <ownedAuthority> elements itself.

So maybe we just recommend to Publishing Registries to find another Full
Registry for managing there Authority id's.

Somebody might look at the CDS example or Full Registries that do
publishing, can we get rid of the elements like <managedAuthority> in that
case, I would suggest that we keep both that way we are clear who is
managing and who is writing/owning the data.

As for your questions below, yes I would assume a Publishing registry could
change/transfer to a different Full Registry if they desired.

Now keep in mind I do believe a authority id can only be managed by ONE
Registry only.  Otherwise report some errors of some conflicts. Same goes
for the <ownedAuthority>.

Now I don't know what to do if the person at the Full Registry refuses to
relinquish  a Managed Authority id if the publishing registry wants to
change :)   Lets hope two people are not that mad at each other.  (there
might be away to transfer authority id management if you can prove your the
owner?)

So yes some more debate in this area is needed.

Cheers,
Kevin

-----Original Message-----
From: owner-registry at eso.org [mailto:owner-registry at eso.org]On Behalf Of
Matthew J. Graham
Sent: 05 April 2005 19:28
To: KevinBenson
Cc: Registry List
Subject: Re: ownedAuthority element


Hi Kevin,

The idea here then is to actually change the meaning of
<managedAuthority>?
Currently we have (if you're lucky since not all registries use the
managedAuthority tag - it's only optional at the moment):

<Registry>
   <identifier>a small publish registry</identifier>
   <managedAuthority>ivo://nvo.apple</managedAuthority>
   <managedAuthority>ivo://nvo.pear</managedAuthority>
</Registry>

which means that the Authority records ivo://nvo.apple and
ivo://nvo.pear reside in this registry and that resources identified
with these authorities can only be published in this registry. I would
say that this really means that this registry *owns* these authorities.

What you are suggesting is that this would become:

<Registry>
   <identifier>a small publishing registry</identifier>
   <ownedAuthority>ivo://nvo.apple</ownedAuthority>
   <ownedAuthority>ivo://nvo.pear</ownedAuthority>
</Registry>

and that there would also be:

<Registry>
   <identifier>the local full registry</identifier>
   <managedAuthority>ivo://nvo.apple</managedAuthority>
   <managedAuthority>ivo://nvo.pear</managedAuthority>
</Registry>

When people wanted to harvest records from the small publishing
registry they would instead harvest them via the local full registry
and it would only ever be the local full registry which harvested from
the small publishing registry?

Would the small publishing registry be free to choose which full
registry managed it? Would it be free to change its managing full
registry if it wanted to, e.g. because it was not being harvested
frequently enough or the verificationLevels that the full registry was
assiging its records were not what it felt it deserved?

I tend to think that the hierarchical model is really only justified if
we had 10000+ registries instead of maybe 10+.

	Cheers,

	Matthew

On Apr 5, 2005, at 2:14 AM, KevinBenson wrote:

> There was something agreed to and thought to be good idea back in
> Boston
> IVOA meeting.  But unfortunately did not make it into the 0.10 schema.
>  The
> idea was to have a <ownedAuthority> element at the same level as
> <managedAuthority> in the Registry type.  Where by <ownedAuthority>
> will be
> the original publishing Registry where the Resource records are
> changed, and
> <managedAuthority> could be at a separate Registry (normally a Full
> Registry).
>
> The reason for this is because numbers of publishing registries are
> starting
> to grow and the current mechanism we have in place a Full Registry
> must go
> around to every Registry (Full or Publish) doing harvests, when in
> general
> it is more practical and beneficial if you could just go to maybe 1 or
> 2
> Full Registries (most likely in a particular
> region/area/country/continent -
> what ever makes the best sense) and pick up all the information.
>
> Now how does the Full(managed Registry) go to a publishing Registry to
> pick
> up there data.  Well I would assume it is also by the harvesting
> interface
> and I would also assume it would be good if the Full Registry (that
> manages
> a Publishing Registry) is down for the moment then another particular
> Full
> Registry may have some logic to go to the Publishing registry to do a
> harvest, but I don't think we should dictate that.  I would rather
> just do
> mostly Full Registry to Full Registry harvests.
>
> For an example (these are just examples):
> HEASARC - may be managed by STSCI Full Registry - NVO
> NCSA - may be managed by Carnivore Full Registry - NVO
> Portsmouth, MSSL, RAL - may be managed by Galahad Leicester Full
> Registry -
> Astrogrid
> Edinburgh(roe) - may be managed by Cambridge Full Registry - Astrogrid
> Tokyo, Kyoto - may be managed by Nagasaki Full Registry in - JVO
> CDS - CDS Publishing Registry
>
> Now instead of Full Registries doing up to 14 different harvests of a
> registry there is just 6.
>
> So is there anything that needs to be better explained?  Does everybody
> still see this as a good idea to implement.  I would think it is still
> a
> good idea and something we should implement when ever we do the next
> release
> of the Registry Type schema.
>
> Cheers,
> Kevin
>