[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: on NAT-PT



i agree.
 what if we do the following:
whatever the incoming response, the dns_alg will forward 2 responses to the
client:
1)if the response is v4 only, then the dns_alg along with the v4 response
will forward a mapped v6 response.
2)vice-versa if the response is v6 only
3)if 2 responses are received then forward them both to the client

a simple functionality in the client may be able to discard the not-required
result. though this will require a bit of effort on the part of the dns_alg,
but is nothing as compared to the effort that excessive translation will
cause.

regards
Anand Thakur
HCL Perot Systems
A-14 Sector-57,Noida
tel ext. - 3257
mobile:9811748512

> -----Original Message-----
> From:	Ronald van der Pol [SMTP:Ronald.vanderPol@rvdp.org]
> Sent:	Thursday, December 05, 2002 6:37 PM
> To:	Thakur, Anand
> Cc:	Christian Huitema; v6ops@ops.ietf.org; juha.wiljakka@nokia.com;
> itojun@iijlab.net
> Subject:	Re: on NAT-PT
> 
> On Thu, Dec 05, 2002 at 13:13:24 +0530, Thakur, Anand wrote:
> 
> > 
> > hi christian,
> > i don't think this a problem in na(p)t-pt, but more a problem in the way
> > dns-alg works. for instance, if we could configure the dns-alg in such a
> way
> > that it does the following:
> 
> Suppose a dual-stack client asks for an AAAA.
> 
> > 1)maintain a table which maps the source ip address of the dns query and
> the
> > type of query ("A" or "AAAA")
> 
> Map that dual-stack client has asked for AAAA.
> 
> > 2)generate a dual-query ("A" + "AAAA") every time a dns query is
> detected.
> 
> Send A+AAAA queries. Suppose you receive A response only.
> 
> > 3)when it intercepts a dns response it should 
> > ->translate the "A" response to "AAAA" response if the original quey was
> > "AAAA" and the response is "A" only
> 
> So, the A response is mapped to AAAA and send to the dual-stack client.
> That's not what you want. You want the dual-stack client to communicate
> over v4 transport with the destination. So, should not send a mapped
> AAAA to the dual-stack client. The dual-stack client should retry and
> ask for A.
> 
> 	rvdp