Multi-conference report: VO and SW

Thu Dec 8 06:21:09 PST 2005

Sorry, I meant to respond to Tony's and Bernard's points also.

So one cup of tea and a slice of (rather nice) chocolate cake later...

Tony:

> c) we also provide DC standard details so the metadata is globally
> available even if the data isn't

This might be said to repair a problem we created ourselves.

On 2005 Dec 5 , at 11.40, Bernard Vatant wrote:

>> Tony wrote:
>> d) no one but an astronomer could possibly interpret our resources  
>> anyway and they'll
> all be 'in the know'.
>
> Yes, I think this is a killer argument. IPTC pushes the use of  
> CURIEs, for similar reasons
> (news agencies will be the only users).

If I understand CURIE's correctly, while IPTC may have been the  
originators, they needn't be the only users.  Hmm.  _We_ could even  
use CURIEs <http://www.w3.org/2001/sw/BestPractices/HTML/CURIE>:

     <html xmlns:ivo="http://www.ivoa.net/ivo-resolver/ivo:">
     ...
     Look at <a href="[ivo:/blah]" >my data</a>
     ...

There: one CURIE-compliant URI which looks like a ivo: scheme, but is  
really an HTTP URL (and you get to pick/override your namespace/ 
resolver).

INFO URIs are interesting, but a bit of a red herring, I think, since  
they're principally intended to name _offline_ resources, and the  
scheme is explicitly not intended to support dereferencing, which are  
the features which jointly justify the new URI scheme.  ivo:  
resources, in contrast, name things which are online, and which you  
do want to dereference.

> Bottom line : SW folks assume an "unique open information world",  
> but the reality is that
> URIs will be used in semi-open (or semi-closed) worlds, or  
> communities of users, inside
> which "universal" identifiers such as http URLs will not be relevant.

There are plenty of URLs on the web that I can't parse, but I never  
look at them, so they never disturb me.  And while it's true that the  
consumers of ivo: URIs would tend to be specialists using specialised  
software, that simply means that we can slightly discount _one_ of  
the new-scheme disadvantages.

My problem with the ivo: scheme is (I now realise) not any claimed  
architectural impurity of a new scheme, but instead the resolver,  
which seems to bring significant costs with marginal benefits.

For the resolver _does_ lock people out.  I don't mean granny, but an  
astronomer who sees an ivo: identifier in a paper.  If this were a  
URL, she could wget it, look at it, `mmm, what's a VOTable?', google- 
google-google, and `aha!, so that's what the VO is!', then go and  
search for a VOTable2FITS converter.  With a small amount of extra  
capability, she could even type it into a web browser and get  
something useful (I'm thinking of GRDDL, but there would be other  
similar possibilities).

The people who grok the VO, or who will willingly play with VO tools,  
do not constitute the whole of the astronomical community.  `Enable  
view-source' is one of the other things that helps create network  
effects, and requiring a resolver gets in the way of that.

Not having a resolver also means that I can write an application in  
whatever (scripting?) language I like, building on massive amounts of  
HTTP support, rather than flaky or partial IVOA resolver APIs or  
protocols.

All the best,

Norman

-- 
------------------------------------------------------------------------ 
----
Norman Gray  /  http://nxg.me.uk
eurovotech.org  /  University of Leicester, UK