Danny Ayers wrote:
> On 8/14/05, Stefano Mazzocchi <stefanom_at_mit.edu> wrote:
>
>>http://wiki.mozilla.org/Mozilla2:Unified_Storage
>>
>>They are clearly looking for a native triple store, but they are
>>confusing it with a embedded relational database.
>>
>>It would kick so much ass if we could convince them to use something
>>like redland instead (and would make my life as a piggy-bank developer
>>*waaaay* easier).
>>
>>Thoughts?
>
>
> Yep, an RDF store within Moz would /calcitare asinus/.
ehmm, s/asinus/asinum/ ;-)
[curiosity: I thought "kicking ass" referred more to the lower back than
to the animal, am I wrong? (scratching head)]
> In straight tech terms I guess it doesn't really matter whether the
> underlying implementation is SQL-DB based or native - there may be
> some advantages in the former, where issues like scalability can be
> pushed to a different (totally solved) layer. But there may be a
> danger hinted in your remark re. confusion. If there was a SQL DB
> underlying an RDF store, there'd be temptation for developers to use
> it at that layer, rather than the more interoperable triple layer.
Right.
> I've a feeling I saw a recent RDF in SQL DB design doc, but can't find
> the link - anyone got a ref? (Coincidentally I just stumbled on
> another case, Steve Harris' updates to 3store [1]). If the Moz guys
> do continue in the direction on the Wiki, it would be a shame for them
> to have to reinvent the wheel.
Why don't we write something in their wiki then? but let's try to be
helpful and avoid not the usual patronizing "thus shall use this because
we know better" semwebby preachy style.
[stefano whissles and looks up]
> Those Wiki pages give me a lot more optimism about RDF in Moz, I've
> seen a few comments in the recent past suggesting movement away from
> RDF with the familiar "too complicated, we can do this with XML"
> argument. They may have more justification than usual, with Moz's
> implementation being begun with the old RDF specs, I get the
> impression of clunky hangovers (i.e. containers). But whatever the
> history, now's the time the RDF capability could *really* be
> exploited.
Yes, I agree. In fact as much as I agree that mozilla's past use of RDF
was combersome and not that useful, it's also true that mozilla's dev
base is one of the few where RDF is a 'required subject', so to speak,
even when not dealing it RDF data directly. Meaning that it's going to
be easier to show its value in such a community then elsewhere (and I
hope things like PB can help there too).
> A couple of avenues come to mind through which the Moz folks might be
> further encouraged. One is SPARQL. It's exceedingly cool - are they
> aware of this?
Unlikely. Another reason to spread some evangelization there.
> The other would to present a prototype based on the
> current build tree as a fait accompli. Anyone got a research student
> with time on their hands? Once a store was integrated, the PiggyBank
> material shows one good route to demoing the potential.
Agreed.
> Redland does seem like a very good candidate, having native capability
> on the different platforms, no need for a separate VM.
Right. Native is key here as nobody would be happy with a 35Mb firefox
just because you need to carry around a JVM for it.
Another option is _at_semantics RDFStore, again native, although I'm not
sure of their SPARQL support just yet (but knowing how much Alberto is
into it I would be surprised if it didn't match it there...I'll talk to
Dirk about it).
Oh, btw, I'm supposed to get together with some of the mozilla folks now
that I'm back to Cambridge, I'll sure mention this.
--
Stefano Mazzocchi
Research Scientist Digital Libraries Research Group
Massachusetts Institute of Technology location: E25-131C
77 Massachusetts Ave telephone: +1 (617) 253-1096
Cambridge, MA 02139-4307 email: stefanom at mit . edu
-------------------------------------------------------------------
Received on Tue Aug 16 2005 - 11:31:55 EDT