> Changes in Linked Data sources (dataset dynamics) and how to deal with it
> are an important and emerging issue [1][2].
>
> As already mentioned, there will be a dataset dynamics meet-up at WWW2010. I
> first planned to register a BoF, but it is still unclear if this is possible
> [3].
>
> I'd now propose to have a break-out session at the W3C LOD Track [4] - as I
> won't be able to make it to WWW, Juergen Umbrich (in CC) would take care of
> the local organisation.
Sounds like a good idea, although I don't know what else might be on the
menu.
So, I just learned of the "dataset dynamics" term and community last
week (from Michael), but I've been thinking about this for many years.
I got inspired and sketched out a design a few months ago, which I think
is pretty good. I've been hoping to return to it some day soon, but if
we're talking about this Thursday, I might was well share the drafts
now.
I think for the use cases I have in mind (large scale, ad hoc, real-time
mirroring of RDF), a key requirement is constant time (per triple) to
apply deltas, including maintaining a secure hash. I was happy to see I
could (I think) meet this with some tweaking of blank nodes. I sketched
both a delta format (gruf) and a subscription protocol (websub), which
are separate.
Rough specs are here:
http://websub.org/wiki/GRUF
http://websub.org/wiki/Spec
-- Sandro