Reply limits exceeded problems

[Havoc Pennington]

On Sun, 2004-08-08 at 16:43, Olivier Andrieu wrote:
> Maybe we could add a convenience function so that message handlers
> (server implementations) can easily send a reply without
> data. Something like this:

I think what we have here is another symptom of the D-BUS API designed
to be the low-level backend for the GLib/Python/Qt bindings. 
Using any of those bindings, the reply should be sent automatically.

> Also, maybe dbus_connection_send() could automatically call
> dbus_message_set_no_reply() on the message ? In other words, make the
> status of the no_reply flag consistent with the way the message is
> sent, i.e. with _connection_send() or _connection_send_with_reply() ?

Right now connection_send() returns the client serial, which is only
useful if you're going to use it to identify the reply later... what
with_reply() adds is that it automatically deals with finding the reply,
timing it out, and so forth.

Which is to say, maybe we should do this but if we did we'd have to drop
the client_serial return.

I'd be more inclined to add connection_send_no_reply() but this is
another convenience function, and currently we have a TODO to delete
some of the existing convenience functions in dbus-bus.h for example...

I'd be a lot more willing to consider convenience functions once the
GLib bindings are working and we know how people are using those... it's
also possible the convenience functions should all be in a "plain C
binding" that's a separate lib from libdbus so everyone doesn't pay for
the extra code.

Havoc