Re: report of the October meeting

On 5 November 2015 at 10:52, Ruben Verborgh <ruben.verborgh@ugent.be> wrote:

> Hi Melvin,
>
> > I've been using rdflib.js for the last 2 years, and im very happy with
> it, in terms of functionality and performance.
>
> Just as a test case, I've briefly set up a performance test:
> https://github.com/RubenVerborgh/JavaScript-Turtle-Parser-Performance
> This repository contains the performance test I already use for the N3.js
> parser,
> plus an adapted version for rdflib.js.
> (I have not translated the specification test, but can do so on request.)
>
> Here are the results for my machine (please verify on your own):
> – N3.js parses 999,999 DBpedia triples in 4,091ms, or 244,438 triples per
> second with 65MB RAM.
> – rdflib.js parses 999,999 DBpedia triples in 37,481ms, or 26,680 triples
> per second with 323MB RAM.
>
> So N3.js parses roughly a quarter of a million triples per second;
> rdflib.js only a tenth of that.
>

Excellent work, thank you!  I've passed this on, hopefully it will be of
interest.


>
> This is not a quality judgement in any way.
> On the contrary, it reinforces my prior belief: we should work together.
> I don't have the time nor ambition to work on a complete RDF/JS library,
> but I do have the expertise to build high-performance parsers and other
> components.
> Others will also have expertise in specific domains and will want to share
> that.
>
> So this is why I strongly believe we need a shared spec,
> which will allow us to write libraries that are compatible with each other
> instead of the competition we see now, which hurts the ecosystem.
> RDF/JS developers, let's unite, and let the Representation Task Force
> be a starting point to combat the current fragmentation.
>

Got it.  My use of the term "hobby project" was not in any way designed to
put down code quality or volunteer work.  On the contrary, the standards
are very high.  What I meant is that some of the work is more like "side
projects", rather than, full time paid work.

So I was just saying MIT has now got funding to hire a couple of full time
people in this area, which wasnt the case before.  The main thing I wanted
to imply is that we're a small team with limited time, compared with huge
projects such as ubuntu etc.

Standardization can be very valuable, but can also take up a good chunk of
time.  Looking at RDF Interfaces it's been improving, but over a number of
years, rather than, a number of months.

Parsers and serializers are something we will all need so that's really a
good thing to work on.  Namely turtle and RDF/XML are RECs and can be
parsed synchronously, so that's a great thing to put in a module.

Of course, that requires an output format too ie quads.   There's not
really that many ways to do quads if we start with RDF 1.0.  Only the
object position may be slightly interesting when adding a language and
datatype, it's probably best to have this in an object, I would say.

JSON LD is a different kind of parsing problem, because it requires an IO
operation for the @context, so is asynchronous.


>
> Best,
>
> Ruben