Universal standard mapping between D-Bus and URI namespaces

[Anders Feder]

tir, 01 07 2008 kl. 20:04 +0100, skrev Simon McVittie:
> Like I said on IRC, I think this isn't a bad idea in principle; while
> I'm not yet convinced of its usefulness, it costs us nothing (we just
> need
> to drop a couple of files in /srv/dbus.freedesktop.org/www/ to
> document
> that some of our URI-space has been delegated to implementors of
> random D-Bus
> stuff).

Regarding usefulness, you mentioned on IRC that you didn't believe that
e.g. ontology alignment of D-Bus interfaces would have much use beyond
so-called trivial cases (i.e. when interface semantics doesn't differ).
This is maybe true to some extent, but surmounting even just the trivial
challenges can get you a long way if you're dealing with interfaces that
differ greatly in syntax.

As an example, you could imagine having an API for sending e-mail
through your desktop PC mail client and another API for sending e-mail
through your mobile phone via Bluetooth. The technical transport details
would differ enormously, but sending e-mail is sending e-mail - the
underlying semantics of the two different protocols would be identical
(most likely modelled on something like SMTP).

Hence, in this case, if you could add a layer that would allow your
application to automatically adapt to syntactically different but
semantically identical interfaces, you would win the ability to send
e-mail by mobile phone with no coding specifically for that interface
what-so-ever.

Theoretically, this adaption could also be achieved by "just" making a
service that map the e-mail-by-phone API into the e-mail-by-desktop-mail
API. But a) this would have to be done for every strange e-mail sending
API out there and b) every one of these services would have to be
included in your application (even those not actually used by a given
user) since the user shouldn't be bothered with having figure out which
service to download and subsequently spend time installing.

When the API map is available as a non-executable description online (in
e.g. RDF), it can be discovered and installed safely without any manual
interaction and without burdening the computer with unneccesary modules
to manage.

Of course, all this would require certain vocabularies (i.e. ontologies)
to be in place first - hence the need for the proposed namespace.

-- 
Anders Feder <lists.anders at feder.dk>