[dccp] Re: RFC - DCCP and NAT

[Ian McDonald <imcdnzl at gmail.com>]

> That NAT box could be several hops away and totally unknown
> to you, therefore I really don't see this as a workable
> solution.
>
You are partially correct there. A couple of thoughts here:
- if you are running 192.168.x.x, 10.x.x.x, etc then the odds are high
that you are running NAT so you could set by default
- the user can set the flag somehow or the OS could even set it if it
fails to establish a connection

Worst case here is that we waste a few bytes of network traffic
sending those options and we didn't need to.

I know that what I propose is not ideal. But my thoughts indicate a
free way forward without needing to replace hardware or flash firmware
(as many/most boxes ares). Hell I had problems a few years ago getting
a ADSL NAT box for under US$100 which supported PPTP and NAT together
and that is at a time when BOTH technologies were common place. Let's
not put barriers in the way if we can come up with a smarter way.
 
> NAT without protocol knowledge is simply impossible without
> giving up checksums and thus data integrity.
> 
I am not talking about giving up checksums in any way, shape or form.
I am talking about how the checksum is calculated and substituting one
piece of data for another.

Having said these things I welcome robust debate around this as I
don't want to do a bad implementation....

Ian