Re: This proposal doesn't imply extra work (was: Requiring DOAP instead of MAINTAINERS file)



On Mon, 2008-01-21 at 14:21 +0100, Wouter Bolsterlee wrote:
> 2008-01-21 klockan 12:54 skrev Mathias Hasselmann:
> > So I understand even less, why you want us to use a file format which as
> > several technical problems:
> > 
> >   - hard to read and write
> >   - redundant with AUTHORS file
> >   - redundant with Changelog, NEWS and FTP
> >   - no support for git or bzr
> > 
> > You want additional information for svn-commits-list, web-sites?
> 
> I've addressed all of these questions in another mail I've just sent. See:
> http://mail.gnome.org/archives/desktop-devel-list/2008-January/msg00169.html
> 
> > You want to provide the service of hosting DOAP files? So keep
> > MAINTAINERS (and AUTHORS, and whatever DOAP related information we
> > already have), add some PROJECT-INFO file that lists the missing pieces
> > of information, and generate the DOAP file.
> 
> This is not extensible without hacking the handling code to use this
> additional/custom information. When using DOAP directly, e.g. FOAF
> descriptions of people may be added directly, as well as properties from
> other namespaces.

With a key file, we can add new fields just as easily as with RDF.  If
we decide we want to add the maintainers' preferred indentation style,
we just start adding:

IndentationStyle=K&R

With RDF, we'd all start adding

<gnome:indentation-style>K&amp;R</gnome:indentation-style>

Except, you know, the gnome: prefix needs to map to some actual
namespace, so we have to decide on that.  And we really ought to make
some sort of schema and put it at that URL.

And changes to code?  We don't have to change the parser for key files.
Any processor that wants to know about the indentation style will have
to do this:

g_key_file_get_value (keyfile,
                      "Project",
                      "IndentationStyle",
                      &error);

But then, if we use RDF, that code has to add this:

some_rdf_lib_get_value (rdf_node,
                        "http://ns.gnome.org/doap-extras/";
                        "indentation-style");

Not a whole lot of difference, except that RDF is more verbose across
the board.  What's nice about RDF is that multiple disparate systems can
begin to understand each other when they all speak RDF.  That makes it a
great *interchange* format.

But a good interchange format is not necessarily a good authoring
format.  If there are systems out on the web that could do cool things
with Gnome if only we provided RDF, then by all means, let's give them
some RDF.

But we don't have to *write* RDF to *produce* RDF.  We can write our
metadata in something simple that maintainers will actually want to use,
which would increase the chances that they'll add more stuff to it.  And
then something (like, say, Pulse) can read that and produce RDF.

And before somebody screams "bike shed" again, let me point out that I
sometimes actually know what I'm talking about.  For the last five
years, I've been watching what happens when you ask people to write
documentation in a convoluted XML format with 400+ elements.  Heck,
people couldn't even get OMF files right.  All of my recent efforts
(including gnome-doc-utils, Mallard, and Pulse) have been focused on
reducing the amount of menial redundant crap writers have to sift
through.

And I'm working on Pulse, so I know how difficult it is to extract
reliable information from package sources.  I'd love to have more rich
and structured data to work with.  But the technical angle is not the
only angle to consider.  People matter too.  No matter how beautiful
your system is to the machine, if it sucks for people, they won't use
it.  And then we'll have, at best, the absolute minimal information
required, and it'll only get updated when it causes a problem.

--
Shaun




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]