Re: Hydra Design Goals: How important is RDF?

I am using the RDF model regularly in my work, but many people today
associate RDF only with RDF/XML and see it as being academic, not useful
for practical problem solving. In that light, it makes the most sense for
the primary documentation and outreach materials to emphasize how Hydra can
make life easier for developers. There can also be supplementary materials
that explain how it is useful to people from the RDF/LOD communities.



On Fri, Oct 2, 2015 at 12:06 PM, Gregg Kellogg <gregg@greggkellogg.net>
wrote:

> > On Oct 2, 2015, at 11:42 AM, Asbjørn Ulsberg <asbjorn@ulsberg.no> wrote:
> >
> > 2015-10-02 16:52 GMT+02:00 Ruben Verborgh <ruben.verborgh@ugent.be>:
> >
> >> We're not disagreeing that the serialization is important;
> >
> > Good! :)
> >
> >> we're talking here about whether or not RDF should underpin it.
> >
> > Are we? I'm certainly not. If that's what I've communicated thus far,
> > my communication skills are terrible. And if so, I'm sorry. With "How
> > important is RDF", I don't mean "I want to remove RDF". What I'm
> > trying to convey is the following:
> >
> > Hydra should have a set of design goals. I think one of these designs
> > goals should be the position RDF has in the technology. In JSON-LD, it
> > was pretty early on made explicit that RDF was not a goal in and of
> > itself (as you just wrote), but a means to a goal regarding
> > self-descriptiveness (and hypermedia controls).
>
> JSON-LD is really useful for developers, but it’s important that Hydra
> work at the model level, based on RDF, and not rely on specifics of JSON-LD
> serialization. In principle, it should be possible to use a Hydra API with
> a different RDF serialization, such as Turtle. The fact that there are a
> number of different RDF serialization formats shows that these can be
> ephemeral; tying to the data model (which, using triples/quads with URIs
> and Literals is pretty simple) allows Hydra to maintain it’s value even if
> another hot serialization format comes along (I hear people talking about
> YAML-LD, for example). IMO, tying an API too closely to the particulars of
> a serialization format is a mistake.
>
> That said, some guidance for serializing using JSON-LD is appropriate, and
> a Primer which is entirely focused on developers using Hydra with a JSON-LD
> serialization will be valuable.
>
> Gregg
>
> > So although RDF enables this in wonderful and powerful ways, it's just
> > a tool. It's a means to an end. So just because something is possible
> > to express in RDF should not make its serialization in JSON
> > unimportant.
> >
> >> It brings us actually to a much more important point:
> >> RDF is crucial, because it enables use cases like Martynas' and mine.
> >> JSON makes it accessible for non-specialized developers.
> >> And JSON-LD is the bridge between the two.
> >
> > Yep.
> >
> >> So how important is RDF? Very.
> >
> > I agree!
> >
> >> It gives specialized client developers what they need,
> >> and—thanks to JSON-LD—also non-specialized developers.
> >
> > Yes! My point exactly! How does JSON-LD accomplish this? In my
> > opinion, by not being designed from an abstract perspective where
> > serialization and such are unimportant details. Instead, JSON-LD
> > acknowledges the fact that syntax is important and it should be made
> > understandable regardless of the fact that it expresses RDF graphs and
> > regardless of whether the consumer of the syntax has heard of RDF
> > before or not.
> >
> >> Without RDF, only one of those groups would be served.
> >
> > Indeed. We should serve both! But I think we need to state this in a
> > design goal, so "unimportant matters" such as serialization can't be
> > ignored.
> >
> > I hope I've made myself better understood now. Just to repeat: I do
> > *not* want to switch out RDF in any way. I think RDF is very important
> > and a huge enabler for both JSON-LD and Hydra. I just don't want it to
> > end up as so important that it lessens the quality of other aspects of
> > Hydra, like serialization.
> >
> > I don't want "You can just map this in JSON-LD" or "To RDF, the syntax
> > doesn't matter" to be arguments against discussing serialization or
> > syntax. I don't think it's possible to ensure this without stating it
> > explicitly in a design goal.
> >
> > :-)
> >
> > --
> > Asbjørn Ulsberg           -=|=-        asbjorn@ulsberg.no
> > «He's a loathsome offensive brute, yet I can't look away»
> >
>
>
>