ietf
[Top] [All Lists]

RE: I-D Action: draft-rosenberg-internet-wait-hourglass-00.txt

2008-02-15 12:41:20
 

-----Original Message-----
From: ietf-bounces(_at_)ietf(_dot_)org 
[mailto:ietf-bounces(_at_)ietf(_dot_)org] On 
Behalf Of Bernard Aboba
Sent: Friday, February 15, 2008 6:21 AM
To: ietf(_at_)ietf(_dot_)org
Subject: Re: I-D Action: 
draft-rosenberg-internet-wait-hourglass-00.txt

Lars Eggert said:

"A big driver for SCTP was for use a signaling protocol. 
Other SDOs are  
using SCTP for signaling in their network architectures, and 
are also  
now introducing NAT functionality at controlled places in these  
architectures. This is why I believe and have argued that an 
IETF BCP  
that documents how to correctly NAT SCTP is the right thing to  
produce. (And, FWIW, DCCP. There's some interest in that as 
well, but  
not such an immediate one as for SCTP.) As a SIP-area person, this  
mode of operation should be familiar to you.

Will this BCP make SCTP available behind a home NAT? Nope. But it  
provides a specification that people can refer to who design network  
architectures that are more tightly controlled than the end user  
Internet, i.e., where people can define and then require 
their NATs to  
have this functionality."

I agree with Lars here -- having a specification is the first step.

There is a milestone in BEHAVE for a SCTP NAT, along the lines of
BEHAVE's milestones (some completed) for TCP, UDP, and ICMP.  We also
have a milestone for a DCCP NAT.

However, I would suspect that clearly specifying how SCTP and DCCP 
work with NAT would eventually make it possible to obtain a home NAT 
supporting those protocols, particularly if implementations were made 
available within the popular distributions (e.g. DD-WRT) on 
which those home NATs are frequently based. 

I am not aware of the status of SCTP NAT code, but I did learn
this week that there is a Linux implementation of DCCP NAT.

On another note, I think it makes a difference whether 
UDP/TCP is combined 
with IP at the waist, or whether UDP/TCP is considered a 
lower layer on 
which IP, etc. can run.  That is, whether we have general NAT 
traversal 
mechanisms which support a wide array of applications, or 
whether we end 
up having to modify each individual protocol.  The draft 
seems to suggest 
the latter approach.  I disagree. 

A simple UDP tunneling protocol sounds useful, too.

There are currently two drafts for running SCTP and DCCP over
UDP.  I look forward to discussing the benefits of a 
per-protocol UDP tunneling approach versus a generic, simple 
UDP tunnel approach.

-d

_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
http://www.ietf.org/mailman/listinfo/ietf