hi David
On 30 Apr 2007, at 12:07, David Reid wrote:
> [ Given it's Queens day I'm not expecting a quick response... ]
>
> What do people think we should do with the DBMS code in RDFStore?
>
> I'm working on simplifying the build process for RDFStore (almost
> done I
> hope) but need to figure out where/how to include/build the DBMS code.
> RDFStore makes repeated assertions that it doesn't rely on the DBMS
> code, but the makefiles seem to suggest otherwise. Additionally, the
> whole of DBMS looks like a separate standalone package that we
> probably
> don't want to include directly as it'll likely have a different cycle
> and audience.
yes they could be distributed separately. The DBMS could be useful as
an alternative to a more traditional SQL DB; to avoid locking and
address concurrency. The layer it provides is basically an networked
hashed k/v store backed by BDB; but any ad-hoc b-tree/hash/mmap DB
could be attached.
> Any suggestions?
to implement the Redland BDB or libb networked storage ?
alternative, as simple k/w store for cached results/queries (non RDF
necessary); or provide a nice RDF layer on top of it to manage the
actual DBMS as a virtual (even though limited) triple hashed store
( this also relates to our internal graph/indexing model too)...
my 0.002 euro...
Yours
Alberto
|