In message <[log in to unmask]>,
Frankie Roberto <[log in to unmask]> writes
>
>Or to put that another way, "is there a variant of 'open data' that isn't
>> Linked Data (i.e. that doesn't require the use of RDF and SPARQL) but that
>> allows the museums sector to achieve workably machine-readable,
>> semantically-interoperable data?". Standards like the OAI-PMH and SRU would
>> be examples of this.
>
>This is a good debate to have (and I still have some doubts as to the
>benefits of Linked Data vs 'open data'), but I should point out, as it has
>been to me, that you don't need to use, or publish a SPARQL endpoint in
>order to 'do' Linked Data - it's just as valid to simply use RDF (the domain
>model) to structure your data, and then to expose this in some form via URIs
>(eg as RDF-XML, RDF-JSON, or some other variant).
>
>It was described to me that SPARQL is 'heavy' Linked Data, whereas simply
>exposing your RDF triples is 'light' Linked Data. And perhaps the work
>that's being done around REST APIs for Linked Data may present a 'medium'
>option...
Well, SPARQL is just a query language for RDF. It's not essential for
the Linked Data project, but it's useful if you want to be able to ask
interesting questions of all this interoperable data we are thinking of
putting out there. Useful because it's a standard - exactly analogous to
the value of SQL for relational data (and with the same proviso: don't
ask your end-users to construct queries using it). Some Linked Data
resources - e.g. Geonames [1] don't offer a SPARQL end-point at all, but
a custom query interface and XML response format. The only problem with
that is that every API is different (which is fine if you only use one).
I think the most important aspect is the use of unique, persistent URLs
as identifiers for the concepts we want to describe, and within the
statements we want to make about those concepts. Three of the four
Linked Data principles quoted earlier address this aspect: only one
mentions RDF and SPARQL. You could just as usefully publish statements
in Topic Map format [2] as in RDF, so long as you used URLs: it would
still be "Linked Data" (at least to me!).
Paradoxically, assigning such URLs to our museum objects is probably one
of the easier bits of the job, since any museum which has registered its
own domain name and has a database-driven web presence and a spare
developer has the tools it needs to publish its collection as Linked
Data. Assigning persistent URLs to your collections objects gives them
an identity which others can then pick up and use, so it's well worth
doing. The real challenge (on the publishing side) is to make statements
about your objects in a way which links to the rest of the Linked Data
web. Very few, if any, collections databases contain information
expressed as URLs.
In my own experiments with the Wordsworth Trust collection [3] I picked
up Geonames URLs on the fly, but couldn't see how to convert any other
data to URLs.
Richard
[1] http://www.geonames.org/
[2] http://www.isotopicmaps.org/sam/sam-xtm/
[3] http://collections.wordsworth.org.uk/object/rdf/GRMDC.C104.2
--
Richard Light
****************************************************************
For mcg information visit the mcg website at
http://museumscomputergroup.org.uk/
To manage your subscription to this email list visit
http://museumscomputergroup.org.uk/email-list/
****************************************************************
|