recursive types work
hp at redhat.com
Wed Dec 29 10:51:47 PST 2004
On Wed, 2004-12-29 at 18:50 +0000, Mike Hearn wrote:
> On Wed, 29 Dec 2004 02:47:54 -0500, Havoc Pennington wrote:
> > They are strictly speaking convenience functions and thus should
> > live in the bindings.
> There's nothing wrong with having some convenience candy in libdbus
> directly: in the ideal world everybody would use bindings, but in the real
> world bindings are typically 'affiliated' with some camp or language and
> some projects would want to avoid that, eg it does not make much sense
> for Wine to depend on glib simply because the raw DBUS API is too much of
> a hassle to work with, so we'd just have to suck it up.
I guess the question is whether we should do some type of "plain C
bindings" (even if those are in the same shared object as libdbus)
For example the "plain C bindings" might also let you register your
objects and implement Introspect() for you. If these bindings were in a
different shared object we wouldn't have to get all angsty about bloat
and could just make them, well, convenient.
It in any case might be worth moving the "convenience API" stuff into a
clearly separate part of the code.
More information about the dbus