These are chat archives for rdfjs/public

21st
Sep 2015
but i think it's a very important topic. short: let's agree on a low level api, not on a internal representation.
Ruben Verborgh
@RubenVerborgh
Sep 21 2015 11:42
@bergos Could you clarify your arguments? An internal representation is very important, and the distinction with "low-level API" might not always be clear.
"internal representation" for me is similar to IRI, Triple, Quad etc. of Jena
Thomas Bergwinkl
@bergos
Sep 21 2015 12:34
@RubenVerborgh the rdf model (graph,triple/quad,node) can be implemented in different packages, because there are good reasons, i described already in linked post. so we have to define the interfaces for these objects, but internal setters and getters can be used to translate the data into the required structures.
Ruben Verborgh
@RubenVerborgh
Sep 21 2015 14:37

@bergos I think we need a more detailed pro/con analysis. I'll sketch some arguments in favor of a common low-level representation format for RDF constructs here. (Note that "internal" is perhaps a misleading term.)
First of all, uniformity means that all modules work in similar ways. This makes development against/of them easy if you are already familiar with another module. Second, translation costs can be high, especially if there are only “formatting” differences (like: library X uses angular brackets around IRIs, but library Y does not).

Then, to reply to some of your points:

can we decide now what internal representation we want in 1-2 years or more?

Yes, Jena has shown this is possible.

but pattern searches in strings will be slow.

Specific use cases will always require specific treatment.

But I guess the important thing is terminology. What I'm talking about is a uniform way to represent RDF-related concepts across modules, and this way should satisfy certain properties we agree on.
Such properties include usability (including: JavaScript-appropriateness) and performance.
@RubenVerborgh "uniform way to represent RDF-related concepts across modules, and this way should satisfy certain properties we agree on" agree, but isn't that the definition of an api/spec for interfaces?
Ruben Verborgh
@RubenVerborgh
Sep 21 2015 15:51
It probably is then—as long as we agree on the contents, I don't mind the term :smile:
Thomas Bergwinkl
@bergos
Sep 21 2015 21:00
ok :smile:
Nicola Greco
@nicola
Sep 21 2015 23:07
brilliant!