[Standards] XEP-0163, v. 1.1pre1

[Remko Tronçon]

> Yes, and Peter's suggestion makes it stay fire'n'forget the majority of
> the time, but takes care of the bad problems you get when the
> configuration starts getting changed (something which as far as I can
> see should be very rare, if ever).

This makes publish+configure rather useless again (except for the race
condition). You have to implement the whole chain anyway in a client,
or you'll be stuck with a badly configured node forever (if it was
configured using a buggy/bad client).

I would like a reconfigure every time you add a configure to a publish
request. I don't really see the problem with that, and it makes both
the XEP, the clients, and the server(?) implementation simpler.

cheers,
Remko