2008-01-21 klockan 12:35 skrev Mathias Hasselmann: > Yes, the Turtle representation looks much more human friendly than the > XML variant. If we really shall directly put DOAP files into our repos, > I'd clearly prefer the Turtle variant over the XML version. Thanks for the heads up. > Still the Turtle representation has all this namespace boilerplate. I > really doubt we should to add yet another layer of boilerplate to our > projects. GNOME uses far too much boilerplate code already... That is because DOAP files (actually, RDF files in general) are "complete worlds" in itself. Namespacing is just a convention to avoid even more useless typing. I might as well use the "foobar:" namespace instead of "doap:". However, I agree with you that for maintainability we should have a convention on the namespace names used. A possible solution is to use a file like project-name.doap.in that contains the actual DOAP content. The build system may then add translated messages (see my previous example) and add the convention-namespaces boilerplate. > Also the rdf:about="http://clutter-project.org/"; attribute seems to be > redundant with the homepage element. Actually, no. The top level entity being described is referred to by a URI for identifying purposes. (Using a blank node is not really useful in case we want to do something useful with the data---compare to a database table without a primary key column). The homepage URI is not a unique identifier for the entity being described, though I used the same URI in my example. The solution would be to introduce a 'namespace' like http://projects.gnome.org/foo that holds all the project names used in the Gnome context. Those are the entities being described in the DOAP files. Note that this URL doesn't need to exist at all; even the hostname is not required to exist. If Olav (or someone else) ever decides to build a Gnome Projects website, we might redirect the url to the actual homepage, in case the project's website is hosted elsewhere. > Well, and using Turtle instead of XML doesn't address those concerns: > - DOAP replicates some information found in the AUTHORS file However, AUTHORS is unstructured text. Not machine-parseable. > - DOAP's release tags replicate information found in ChangeLog, > NEWS and on FTP servers However, ChangeLog and NEWS messages are pretty much unstructured text. Not machine-parseable aither. > - DOAP's repository support is limited to SVN, BK, CVS and Arch. That is easily extendible by introducing a new namespace with anything we might need in addition to the DOAP-provided properties and classes. Gnome might even add additional properties to encode information such as "is-part-of desktop-suite" and the like. mvrgr, Wouter -- :wq mail uws xs4all nl web http://uwstopia.nl never lived never lived :: i'm in love -- black rebel motorcycle club
Attachment:
signature.asc
Description: Digital signature