[e2e] Internet Draft and survey on P2P in the presence of NAT

[David P. Reed]

At 11:19 AM 4/9/2003 -0700, Christian Huitema wrote:
>When it comes to NAT traversal, Bryan's survey omits an important
>reference, the UPNP "Internet Gateway Device" specification, available
>at:
>http://www.upnp.org/standardizeddcps/igd.asp. According to a recent
>empirical survey, this specification is supported by about 60% of the
>NAT now sold in the US, and 99% of the NAT sold in Japan.

The problem with the UPNP IGD is that you cannot easily implement it for 
non-Microsoft/non-Intel platforms, because it is embedded in a proprietary 
set of specs.

Or maybe I should say that I personally cannot find a simple description 
that enables one to write just the IGD code in my favorite language 
(Smalltalk) or, for example, Python or Java.   Intel claims to have Linux 
code for UPnP, but look closely at the licensing restrictions.

Yeah, it might be possible to build all million lines of a full UPNP 
implementation after reading all the specs.   But that's like saying I can 
run a 3rd party windows app on Linux because I can emulate a Pentium and 
run windows on the emulator (oops, can't emulate the proprietary parts of a 
Pentium that Intel shares with Microsoft, if there are any, as I suspect 
there are) .

Let's keep techniques like NAT traversal open and not entangled with 
semi/pseudo-proprietary protocols.