RE: Pre-release of my partial TPF implementation

On 15 Okt 2014 at 09:36, Kjetil Kjernsmo wrote: 
> I released a pre-release of my TPF implementation last night:
> https://metacpan.org/release/KJETILK/RDF-LinkedData-0.69_03 It is part
> of my RDF::LinkedData module, a module that has been out there for a few
> years already, doing mostly the 303 dance on data in a triple store, but
> also integrates an optional SPARQL Endpoint, CORS headers, VoID
> description, ETags, RDF-based packaging, etc.

Fantastic!


> My TPF implementation has a 45 subtest test suite, which passes for me,
> passes in Travis CI, but still has some problems on FreeBSD, apparently.
But
> hey, I said it was a pre-release! :-)

:-)


> I think the implementation experience was quite good, it didn't take me
more
> than 3 days of work, but the code got somewhat lengthier than I had
> anticipated. It doesn't do paging, mainly because I don't know if it'll
save
> us anything on the backend (but also due to a vague idea that if you need
> paging, you should have been more careful about the selection in the first
> place), but that's about the ambition I had for the next full release.

Fair enough. Everything works without paging as well but you might run into
limitations without paging.


> I think the main pain point is the parsing of parameter values:
> https://metacpan.org/source/KJETILK/RDF-
> LinkedData-0.69_03/lib/RDF/LinkedData.pm#L303 I don't quite like the
> look of my code there, since I use regular expression based heuristics
> to do it. I see this play into the IriTemplate discussion that you've
> had. I have not been involved in that, I must admit.

My Perl is a bit rusty but the code looks correct. The only thing I noticed
is that the underlying RDF library is apparently still based on RDF 1.0 (or
handles this internally) . In RDF 1.1 (language-tagged) strings have always
have a datatype of xsd:string (rdf:langString) [2].


> Though they aren't pretty, I tend to think that a regular expression would
> be the way people would implementing stuff like that anyway, so how about
> defining it in term of a regexp? Just throwing it out as an idea. :-)

We can certainly add a regex for, let's say Javascript, in the spec but I
wouldn't base the definition on a regex. If we want to make it more formal
(A)BNF is probably better.


> Another point is that RDF::LinkedData is in Debian (since Wheezy and
Ubuntu
> 12.04 LTS). The Debian freeze is imminent, but if I release the module
this
> week, it is quite likely to be in the next Debian Stable release. So, my
> question is "should it"? TPF is pretty straightforward, so I am pretty
sure
> I've nailed the essentials. Also, even though this is in stable, it is
> nothing that prevents users from upgrading just the module later.
> However, I certainly wouldn't want my quirks to impede the progress of
TPFs
> by putting a partial implementation of a spec that isn't finalized into
> something as widespread as Debian Stable. So, what do you folks think
about
> that? I can take the truth. :-)

I personally would be fine with this release but Ruben should know best.


Keep up the great work (and keep us posted),
Markus


[1]
https://metacpan.org/source/KJETILK/RDF-LinkedData-0.69_03/lib/RDF/LinkedDat
a.pm#L322
[2] http://www.w3.org/TR/rdf11-concepts/#section-Graph-Literal



--
Markus Lanthaler
@markuslanthaler

Received on Wednesday, 15 October 2014 09:48:04 UTC