Re: Recommendation: Plan for migration

Hi everyone,

I'm just reading all this thread at once, and I think I like the evolution. At the beginning it sounded (to my non-native English speaker ears) like recommending to come with a global plan, which should beimplemented by all.

Now it seems it would rather be geared towards identification of various steps, or even independent realizations. And common problems, for which solutions could be shared. If that's the case, it would be much better!

In fact perhaps this matches a bit my own experience at Europeana: we're using RDFa, but the motivation for this specific feature was quite disconnected from any global LOD project.

Cheers,

Antoine


> Ed, these are good points. I wonder if we can get this into the document as an area to be studied. There are obviously issues and difficulties with both (all?) options and there really isn't a way to know today what LD might evolve into ten years or so down the road. Perhaps we can say that there are lots of ways to get to LLD and that it isn't something that has to happen all at once. (On a recent panel I was on we came up with the Y2K analogy: there is no Y2K drop dead date for LLD and we don't have to all move in the same direction at the same time.) However, we think it should involve planning where possible. The recommendation would be able to refer to less-disruptive options like an interface layer over current systems as something to consider as a starting point.
>
> One thing that to me colors all of this is that we really are coming to the end of the utility of the MARC format, for reasons unrelated to LLD but because of its own limitations. Any change to our primary data carrier needs to be forward-looking. This therefore puts us in one of those "opportunity" moments. I personally think that we could replace MARC without embracing LLD but we should at least adopt some of the primary principles that have come out of the semantic web -- not because they are semantic web but because they are simply a good idea for data management. For example, using identifiable things rather than text strings; defining your properties in a machine-readable way; etc.
>
> kc
>
> Quoting Ed Summers <ehs@pobox.com>:
>
>> I like the idea of emphasizing Linked Data as an option for publishing
>> data, and de-emphasizing discussions around converting all our MARC
>> data over to RDF, followed by powering down the legacy MARC-based
>> systems.
>>
>> For example, there is a lot of iterative work that ILS vendors could
>> do now to layer RDFa into existing OPAC displays, to make the data
>> easier to scrape, and for search companies to index. This would
>> necessarily imply thoughtful use of URIs as well. But it would not
>> mean that their back end data store would need to be thrown out, and
>> all their MARC data converted over to RDF.
>>
>> //Ed
>>
>>
>
>
>

Received on Tuesday, 3 May 2011 20:53:21 UTC