Binary RDF

[Kjetil Kjernsmo]

On Friday 19. August 2011 22:50:18 Gregory Williams wrote:
> I won't argue that there would be big benefits to refactoring the system
> in this way. Patches welcome, as always, but I certainly don't have
> time to do it in the foreseeable future. 

Yeah, the Moose::Role refactoring is a very interesting direction, and one 
that we did touch briefly at the hackathon. However, I think this is an 
undertaking that would require a larger community than we currently are, 
but I am still hoping to discuss it if we are able to mount another 
hackathon.

> 
> That being said, I'm not sure such a refactoring has anything to do with
> my skepticism of the HDT format being discussed. 

I agree, I see this as mostly orthogonal (at least 80 degrees :-) ), 
especially in light of that the benefits could be reaped by a clever 
implementation of get_statements, as far as I can see.

> My big concerns 

[snip stuff I can't comment on]


> The document contains details for algorithms working directly over the
> HDT format, and shows that certain triple access patterns can be
> directly answered, but I'm not convinced that this is all that
> valuable.

Actually, I think it is. At Opera, I was argueing that the browser should 
support querying triples where the subject was the page currently being 
visited, and from there, it could explore whether it was more relevant 
stuff that the user should know about (it is actually one of my original 
motivations for getting into semweb, looking up whether the page you're 
viewing has critisism elsewhere). The store would thus look only triples of 
a certain subject, and this would seem well suited for that.

Anyway, interesting to hear Axel's elaboration too, I welcome efforts on 
binary serializations of RDF and as a member submission, I think it is an 
interesting start.

Cheers,

Kjetil
-- 
Kjetil Kjernsmo
Ph.d Research Fellow, Semantic Web
kjetil at kjernsmo.net
http://www.kjetil.kjernsmo.net/