Hi Brian,
I apologize for not responding sooner, your mail went into the
ietf@xxxxxxxx bucket, which generally gets ignored. (Thanks to Ray
for pointing this out.)
On 17 July 2013 11:14, Dickson, Brian <bdickson@xxxxxxxxxxxx> wrote:
> o What domain name is to be used, in the SRV query ("_stun._udp.domain"
> or "_stuns._tcp.domain" -- what value of "domain" is to be used in the
> query)?
>
> Without that being specified, implementers/users have nothing specific to
> try, and without that, I do not see how STUN could be used at all.
>
> Similarly, recommending that STUN be provisioned (section 4.6) leaves the
> question of "under which domain", unanswered and ambiguous.
STUN does define a discovery mechanism, but that mechanism depends on
configuration of an input domain, as you rightly identified. However,
no local discovery method has been defined for STUN, partly because
that can be somewhat self-defeating, depending as it does on the
server being outside of any local access.
The normal practice for STUN is, as you fear, configuration (see for
example, the information at http://stunserver.org/). On the upside, a
STUN server is relatively cheap to operate. Such a server can live
almost anywhere in the network.
--Martin