[Patch] Moving FD passing into the header

[Colin Walters]

On Wed, Jun 16, 2010 at 10:29 AM, Marcel Holtmann <marcel at holtmann.org> wrote:
>
> I have to agree here. Why are we trying to treat file descriptors
> differently? I don't see the point in here. I like having them part of
> the signature and part of the introspection.

Because changing the type system impacts *everything* in a huge way.
All bindings (and there's a lot of them) would need to be carefully
updated to ensure that they handle the new type.  And there are a lot
of tricky cases there, like variants.

I just don't expect that much code to be using file descriptor passing
that it's worth the pain that would entail.

Now - as far as documentation goes - I agree this is an important
issue for dbus; people have been sort of solving this by various
approaches such as hand-writing the introspection XML and merging that
in with an external <doc:doc> namespace or whatever.  And if you're
doing that, you can fairly easily document in them "This function
returns 1 file descriptor" or the like.

It would be quite nice if we had a central registry for well-known
services combined with a doc format; there are some random things in
/usr/share/dbus-1/interfaces, but...it needs some standardization.
(If it was, then tools like d-feet could also display docs inline).