W3C home > Mailing lists > Public > ietf-discuss@w3.org > December 2002

Re: Application protocols and Address Translation

From: Graham Klyne <GK@Ninebynine.org>
Date: Mon, 02 Dec 2002 17:54:54 +0000
Message-Id: <>
To: Tony Hansen <tony@att.com>
Cc: discuss@apps.ietf.org

Good point.

More generally, NATted systems only really "work" when they make only 
outgoing connections.  (Yes, you can define a "pass-through", but that's a 
horrible hack, and only lets you have one receiving endpoint.)  If one 
looks to the coming generation of peer-to-peer protocols, I think that's 
going to be a significant problem.

Another area that has attracted some interest of late is asynchronous 
notification protocols.  Currently, it seems that one must keep an outgoing 
connection alive in order to receive notifications (e.g. like SMS to mobile 
phones -- and look how mobile phones have taken off in recent years 
compared with Internet access).  Long-lived (TCP) connections get disrupted 
by address changes if the ISP decides it needs to reallocate the address 
used.  How are such applications to operate between pairs of systems that 
are both sited behind address translation?

Where this all leads, I think, is that the worst thing about NAT is that it 
hinders the deployment of new applications.  One of the reasons the 
Internet was able to take off the way it did was that individuals and small 
groups could dream up new applications and deploy them easily.  By 
hindering that, NATs threaten to cause stagnation of Internet development, 
and loss of all the potential ensuing business activity.  (Hey, maybe 
that's why the dot-com boom went bust -- because NAT-like deployment was 
sufficiently advanced that companies could no longer deploy the kinds of 
new service that people really found attractive?)


At 12:27 PM 12/2/02 -0500, Tony Hansen wrote:
>Patrik Fältström wrote:
>>On måndag, dec 2, 2002, at 16:06 Europe/Stockholm, Brian E Carpenter wrote:
>>>The question is, what are the applications-specific arguments
>>>against translated addresses? What are the brokenness conditions
>>>caused by translated addresses (and the associated statefulness)?
>>FTP, SIP and DNS are three protocols which have problems. Or not really 
>>SIP but anything which uses RTP and is launched by an SDP specification.
>Make that any protocol that negotiates for a separate peer to peer 
>connection, and passes IP addresses as part of that negotiation.
>         Tony Hansen
>         tony@att.com

Graham Klyne
Received on Monday, 2 December 2002 15:49:18 EST

This archive was generated by hypermail pre-2.1.9 : Tuesday, 24 February 2004 19:46:24 EST