Re: [sipcore] WGLC: draft-ietf-sip-dual-stack

[Adam Roach <[e-mail]>]

[as individual]

On 3/29/16 13:19, Dale R. Worley wrote:
> As far as I can tell, the goal of this section is to make it clear that
> if SRV records translate a SIP domain into (e.g.) two SRV records with
> different priorities, and one SRV record's target name only has IPv4
> addresses and the other SRV record's target name only has IPv6
> addresses, then which family of addresses is attempted first is to be
> determined by priorities of the SRV records.  This being the process
> described in section 4.2 of this draft.

I think it's somewhat more general than that, though. if you look up an 
SRV record and get multiple A and multiple AAAA records, and those 
records in turn contain multiple literal IP addresses, then you apply 
the SRV priorities first and the RFC 6724 processes second.

I think an example might help.

Let's say I look up an SRV record (_sip._tcp.example.com) and get back 
two entries:

_sip._tcp.example.com.  300 IN  SRV  10 1 5060 sip-1.example.com.
_sip._tcp.example.com.  300 IN  SRV  20 1 5060 sip-2.example.com.


So then I grab the A and AAAA records for sip-1.example.com and 
sip-2.example.com:

sip-1.example.com.    300  IN  AAAA  2001:0db8:58:c02::face
sip-1.example.com.    300  IN  AAAA  2001:0db8:c:a06::2:cafe
sip-1.example.com.    300  IN  AAAA  2001:0db8:44:204::d1ce

sip-1.example.com.    1800  IN  A  192.0.2.45
sip-1.example.com.    1800  IN  A  203.0.113.109
sip-1.example.com.    1800  IN  A  198.51.100.24

sip-2.example.com.    300  IN  AAAA  2001:0db8:58:c02::dead
sip-2.example.com.    300  IN  AAAA  2001:0db8:c:a06::2:beef
sip-2.example.com.    300  IN  AAAA  2001:0db8:44:204::c0de

sip-2.example.com.    1800  IN  A  192.0.2.75
sip-2.example.com.    1800  IN  A  203.0.113.38
sip-2.example.com.    1800  IN  A  198.51.100.140


What I believe we're saying is that I will process these addresses in a 
batch, altogether, using RFC6724 section 6 rules:

  * 2001:0db8:58:c02::face
  * 2001:0db8:c:a06::2:cafe
  * 2001:0db8:44:204::d1ce
  * 192.0.2.45
  * 203.0.113.109
  * 198.51.100.24


And then (but only then, after I've exhausted that entire batch of 
addresses), I'll process these addresses in a batch, altogether, using 
RFC6724 section 6 rules:

  * 2001:0db8:58:c02::dead
  * 2001:0db8:c:a06::2:beef
  * 2001:0db8:44:204::c0de
  * 192.0.2.75
  * 203.0.113.38
  * 198.51.100.140


... and, most importantly, I won't mix the batches together.

/a