Where will we end up? - Re: Toward easier RDF: a proposal

I’m on holiday in the sun, with very bad connection, so I will indulge in some reflection.

Dave De Roure and I have exchanged a mantra for years now:
“Putting the Web in Semantic Web”
I think it hasn’t happened, and I worry that the Web part of Semantic Web *still* does not seem to be much of an objective for many people in the Semantic Web research community.

Many years ago (during the AKT Project) I was musing on how we would know when or if our Semantic Web research was a success.
I suggested that the realisation would not be our "Semantic Web", with RDF and all that lovely stuff.
But we would know we had succeeded when we were sitting looking at the mainstream stuff around and saying:
“That’s our Semantic Web, but they don’t call it that”.
And even adding: “Not fair!!!”.

This was based on historical observation of other things I wondered about.
For example:
What do Simula researchers think of OO all over the place?
What do SGML researchers think of XML?
What do Functional Programming researchers think of lists and λ-expressions available all over popular languages, and pattern matching and ADTs, even?
What does Tim (or Hypertext people) think of the Web? ;-)

None of these things were the sole preserve of those topics at the time - ideas always come out of a context, but they were important milestones.

On a personal level, I got out of pure FP research rather than get involved deeply in Haskell - I could see that lots of stuff about side-effect free programming and encapsulation etc. were now widely accepted.
But it just seemed that FP research people wanted to take over the world and all application domains with that stuff (monads, anyone?), and we were moving from something of beautiful simplicity to something that was too complicated for people to bother to understand and adopt.
Why on earth would you want to write a “Purely Functional Operating System”?
Yes, if you want to push the limits of understanding for research:- but as a useful topic to sell the ideas of FP, it is likely to be a waste of time.
After a short description of pure FP, anyone can tell you it is unlikely to be a great technology to write an OS kernel - yes, some of the ideas are useful, but pure FP tech? No?
That may be a digression, but I do feel a sense of relevance to this discussion, as there are parallels with SW and our discussions here.

And this is the nature of the research process.
The essence and real wins of research that are successful get adopted into the mainstream - and usually not the vehicle that was used to extract the essence, and work out what the best bits are.
The research vehicle is really important, it allows researchers to explore the subject, even to apparently extreme levels, and to continue to do it after things have moved to the mainstream, but it is not the delivery vehicle.

By this measure, the SW research has not been a success.
Or has it?
What might we be looking for as essence?
Much of the stuff we have discussed.
Things like common vocabularies, globally unique identifiers, schema-less storage, well-founded graph models, knowledge exchange protocols, inference rule description.
If we look around, we see a lot of that stuff has grown immensely in the last decade.
Some of it even because of the Semantic Web activity, and even by the same people.

What does all that mean in the context of David’s call to arms?
I’m not sure what - this is just my indulgence.
It could mean that we declare Semantic Web research a success because of Schema.org, NoSQL/GraphDBs, Google/Facebook Graphs, and maybe other things - and move into the mode of continuing to push the boundaries and see what else comes out.
But I don’t think so.
For a start, I don’t see where the RDF model has got picked up.
There isn’t a LARP that I know of, so that RDF is not an easy choice even for appropriate applications.

What it may mean is that we should be trying to make chosen bits of the Semantic Web more useable and in particular, useable as part of other technology choices for developers.
And losing things, such as “difficult"/existential bNode usage or even bNodes themselves, and other things that we aficionados think of as “good”, may be the right thing to do, and a price worth paying for us all to get the benefit of being able to use the technologies easily.
And unbundling, so that bits of the ideas can be migrated to other places (Schema.org?).
And then we might even be able to point with pride at things being very widely used and say we were part of that.
<puts head above parapet>
There was a time when I saw the Linked Data initiative as this split; but Linked Data hasn’t managed the direction David is talking about - possibly because it didn’t lose the Semantic Web people.
The LOD workshop contributions looked like they could easily be Semantic Web papers.
So we started the COLD (Consuming) workshops, but immediately they looked pretty indistinguishable from the LOD papers.
It seemed to me there were very few people contributing who were actually consuming data from the wild from multiple heterogeneous or even homogeneous LD sources and building significant applications with it.
And that is meant to be the Semantic Web/Linked Data sweet spot - not even a Purely Semantic Web Operating System.
<puts head above parapet />.

I think David may be trying to make this research/adoption thing happen, of course.
But maybe this will just repeat the Linked Data experience?

If you got this far, I remind you I did warn you I would be (self-)indulgent. :-)

Received on Wednesday, 28 November 2018 13:57:35 UTC