Yes it should, but first, it needs to implement its draft version more sanely.

LOOK MA, I DID IT.

This looks fine. I've got a couple of nitpicks, but neither is a blocker.

+      const gchar *client_name = g_value_get_string (va->values + 0);
+      const GPtrArray *filters = g_value_get_boxed (va->values + 1);
+      const gchar * const * cap_tokens = g_value_get_boxed (va->values + 2);

I could suggest using tp_value_array_unpack here? I don't know how much clearer it would actually be.

  if (!announce_self_caps (self, &error))
    {
      gabble_capability_set_free (before);
      dbus_g_method_return_error (context, error);
      g_error_free (error);
      return;
    }

  if (!gabble_capability_set_equals (before, after))
    {
      _emit_contact_capabilities_changed (self, base->self_handle);
    }

I'm tempted to say that the former block should be inside the latter — why would we want to publish updated caps if they haven't changed? I know this isn't a regression in your branch.

(In reply to comment #3)
> +      const gchar *client_name = g_value_get_string (va->values + 0);
> +      const GPtrArray *filters = g_value_get_boxed (va->values + 1);
> +      const gchar * const * cap_tokens = g_value_get_boxed (va->values + 2);
> 
> I could suggest using tp_value_array_unpack here? I don't know how much clearer
> it would actually be.

I'd rather just keep it the same tbh as it'll be the same as gabble, not sure how much of an argument that really is.. anyway, yeah, deal widdit.

> I'm tempted to say that the former block should be inside the latter — why
> would we want to publish updated caps if they haven't changed? I know this
> isn't a regression in your branch.

Yeah you're right, I've fixed this. See the latest patch on my branch then.

OH, and I've added capabilities-set and caps-channel-manager to the plugin API. See this branch on top of this one:

http://cgit.freedesktop.org/~jonny/telepathy-salut/log?h=plugin-caps

both look fine!

le done.