[DAS2] Ontology URIs (was RE: types.rnc)

[Andrew Dalke]

Chris:
> Andrew
>> We could.  When making the RNC I left the "SO:" prefix out
>> deliberately in the number, leaving "0000147".  The reason was to
>> be very insistent that it was SO and only SO that could to there.
>
> This seems overly defensive. It would seem cleaner to use the same ID 
> scheme throughout

What I wrote was

> The sequence ontology (SO) is widely used but its identifiers are not 
> URIs. The 'so_accession' attribute contains the SO accession number 
> without the leading "SO:", as in "0000316". Note that the leading 
> zeros are important. This field should be interpreted as an opaque 
> string. (XXX should this be "0000316" or "SO:0000316"? I prefer the 
> latter.)

The "XXX" in the spec mark places where I'm hoping for feedback.
So far I haven't received any.


Chris:
> Andrew:
>> But if we go for systems with no default resolver, why not use
>> LSIDs?
>>
>>    url:lsid:biodas.org:go:0000147
>
> LSIDs uniquely identify sequences of bytes. The sequence of bytes in 
> the record GO:00000147 may change although the universal it refers to 
> does not

LSIDs have concrete objects and abstract objects.  The abstract object,
if resolved, only returns metadata.  This would be an LSID for an
abstract object.

If we have a so_version and so_accesssion, etc. as attributes then
we could identically have an LSID referencing an abstract object.
It makes no difference to clients and for the spec is promotes the
push towards URIs.

					Andrew
					dalke at dalkescientific.com