[Standards] XMPP Registrar: Registration policy

Peter Saint-Andre stpeter at stpeter.im
Thu Dec 29 21:50:26 UTC 2016


On 12/29/16 9:39 AM, Sam Whited wrote:
> Hi all,
>
> I recently proposed a change [1] to the way the XMPP registries work to
> require that all registries have a ``registration policy" which would
> determine how entries are added to the registry (eg.
> first-come-first-server might mean that anyone could simply add an
> entry, but spec-required might mean that some normative text that
> explains how the registry entry works and what it's for be written
> first).
>
> Peter suggested that I bring this to the list, and it does seem worth
> discussing, so here it is. Peter's view was that:
>
>> The IETF has been moving toward "First Come, First Served" in most
>> cases, and I don't see a compelling reason for the XSF to move in the
>> opposite direction by adding more formality to our processes.
>
> In my mind the only real distinction that's necessary is "should entries
> in this registry require a normative description, or is this just a
> non-normative list that can be safely ignored". However, if the registry
> is only intended as a way to make sure your custom namespace, feature,
> option, etc. doesn't conflict with other peoples usage (but without
> guaranteeing any compatibility if you choose to implement things in the
> registry already in use by other people), then I agree with Peter, it's
> easier just to keep it first-come-first-served.

Folks at the IETF used to treat the process of adding an entry to a 
registry as a quality check: if someone needs to publish an RFC in order 
to add an entry, bad entries will be prevented. In practice, this 
approach has proved less useful than you might expect. Furthermore, 
until recently the XSF has added entries to our registries only after 
the relevant specification has advanced to Draft (for the most part - 
there are exceptions). Personally I'm in favor of adding more process 
only if there is a clear need for it, and so far I don't see a clear need.

Peter



More information about the Standards mailing list