Re: [BEHAVE] Opsdir Review of draft-ietf-behave-turn-uri-03.txt

[Marc Petit-Huguenin <petithug at acm.org>]

Margaret Wasserman wrote:
> 
> Hi Marc,
> 
> On Oct 30, 2009, at 3:36 PM, Marc Petit-Huguenin wrote:
>>>
>>> (3) But, how could you transform "turn:example.com" into an A/AAAA
>>> record query? Where do you get the host name?  It doesn't make sense to
>>> just do an A/AAAA query on "example.com" when you are trying to find a
>>> specific server...
>>
>> $host example.com
>> example.com has address 192.0.32.10
> 
> I didn't say that you _couldn't_ send a query for example.com, just that
> it wouldn't
> make sense to do so...
> 
> Are you expecting that administrators will accept a requirement to run
> their TURN
> servers at an IP address returned for their domain, in order to maintain
> their ability to
> move to DNS SRV records or S-NAPTR later without changing the URIs on their
> end-user devices?
> 
> Isn't it more likely that they will want to name their TURN servers
> something like
> turn1.example.com, or something like that?

Or better turn.example.com.

A provider starts with A/AAAA RR, so turn.example.com is an FQDN, then later
they can convert it into a domain name by adding NAPTR/SRV.

> 
>>> This was what I was dancing around in an earlier post, when I indicated
>>> some confusion about whether the <name> portion of the URI was intended
>>> to be a FQDN or just a domain...  The specification seems to have some
>>> ambiguity in this area.
>>
>> "FQDN" is never used in the draft, so I do not understand why you are
>> referring
>> to this.  A <host> can be a FQDN and a domain name at the same time.
>> Even "com"
>> could be a FQDN if someone adds a A record for it, why not?  As there is
>> absolutely no way to distinguish between a FQDN and a domain name, I
>> do not see
>> the point of using this terminology.  If you really want to think in
>> term of
>> FQDN and domain name, then the algorithm is to take whatever is in the
>> <host> if
>> it is not an IP address, use it first as a domain name and if the
>> NAPTR/SRV
>> query fails, use it as a FQDN.
> 
> I am not attempting to argue about the term FQDN vs. the term domain
> name...  I
> am arguing that if you won't want to use the same <name> value for the DNS
> SRV lookup as the A/AAAA in most cases.  So, I am not sure how well the
> fallback
> you have described will work in practice.
> 
> Are there cases where people do something like this today and it works?

SIP (RFC 3263).  Yes it works, I used it for the VoIP service I developed for my
former employer.

-- 
Marc Petit-Huguenin
Personal email: marc at petit-huguenin.org
Professional email: petithug at acm.org
Blog: http://blog.marc.petit-huguenin.org