Being interested in NameOwnerChanged for a namespace, not one name
marcel at holtmann.org
Fri Sep 17 16:59:59 PDT 2010
> > We should consider the optimizability of whatever we do. I think
> > there's already a bug or an issue about O(n) searches through match
> > rules on every message.
> I'm happy to report that I improved the daemon about a year ago to index
> match rules by message type and interface. :)
> That said, I don't think that globs are necessary. I really can't think
> of any general use-cases for wanting to match anything more than a
> prefix of bus names (or potentially interface names, for the first
> argument of the new org.freedesktop.DBus.Properties.PropertiesChanged
> signal). Matching a prefix is just prefix_length <= actual_length &&
> memcmp(); glob matching is way more complicated, so without a
> justification I'm inclined to pass on that.
for the bus name, maybe not. For the object path, it might makes sense.
I personally think that this design of waiting for bus names, object
path or interface names with a well defined prefix is wrong anyway. That
is why I have never resorted to this. I am using an active registration
process where the client has to register with the central server. That
way it learns the bus name of the client and can track its lifetime.
If we are trying to fix Telepathy/MC then it might be better to actually
fix it inside Telepathy than hacking around it in D-Bus protocol. Just
my 2 cents here.
More information about the dbus