[Debwebid-discuss] rdf coordination

[chrysn]

On Wed, Aug 12, 2015 at 04:58:11PM +0100, Iain R. Learmonth wrote:
> Ok, so the service I have running at rdf.debian.net is going to switch to
> be entirely just minting URLs. The dataset got waaaaay too big for there to
> be a copy of the full dataset stored in RAM so these need to be minted
> dynamically (unless someone wants to volunteer a large server).

We might be using different terminology here, but in my understanding
the minting process primarily means deciding on a URI for a thing (or
assigning a meaning to a URI), which not even necessarily requires there
to be a server backing that URL (as we want to be nice participants of
the semantic web).

> I have some semi-working code to act as a framework for this. I plan to also
> move this to be meta.debian.net and not rdf.debian.net as it's been pointed
> out that FTP doesn't really exist anymore and so ftp-master doesn't make
> sense and CDs don't exist anymore and so the CD team doesn't make much
> sense. Naming the service after a technology is probably not the best idea
> and so once I'm a DD (shouldn't be long now) that new service will be going
> up.

I'll chat up rhonda on that topic too, afaik she's been having an eye on
having cool URIs that don't change(TM) on the web frontend for some
time.

Actually, we currently have two models that describe roughly the same
things, for example [1] and [2]; I'd leave out the topic of verbs for
this thread. We should make up our minds about canonical URIs for those
things. Not that it'd be *essential* for using the data (when both state
that their doap:homepage is [3], they can be smushed), but it'd make
things easier.

I'd prefer a style that neither of the current approaches use, but is
closer to the PTS version's name:

http://packages.debian.org/src:arandr#project

It's as badly dependent on the transport technology as any other URL we
could dereference (for what I might consider a design flaw in URI
design), but the hostname is semantic, we can tell apart the page from
the thing with the fragment identifier without *necessarily* going
through a 303 to make sure the thing and its description don't smush.
I'm not a big fan of having the URIs in a separate namespace from what
we use for describing the things in regular life (which would apply to
both rdf. and meta.) because we're talking about the things here,
there's nothing meta about them. (In the Plato cave example, that would
be like saying "shadow of chair" when talking about the chair).

[1] http://rdf.debian.net/project/arandr
[2] http://packages.qa.debian.org/arandr#project
[3] http://packages.debian.org/src:arandr

> > the next thing i'd like to tackle is debtags, which should really be
> > available in rdf.
> 
> If you can make the data available in UDD and provide queries for getting
> these and work out what schemas to use, then I can integrate these into the
> meta.debian.net URL minting service.

From the schema[4] I'd say that the information is already inside UDD.
The metadata about the tags is not available, though.

[4] https://udd.debian.org/schema/udd.html

> Even better, if you can provide code that grabs the queries from UDD and
> returns an rdflib graph with all the relevant information in it, then I can
> just drop that in.

I'll have a look.


Summarising, the question I'd like feedback on in this thread is: What
is the most suitable URI, for example, for "the arandr source package"?


Thanks for your response.
chrysn

-- 
To use raw power is to make yourself infinitely vulnerable to greater powers.
  -- Bene Gesserit axiom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/debwebid-discuss/attachments/20150812/8b2eff54/attachment.sig>