[e2e] What should e2e protocols know about lower layers?

[Joe Touch]

Erik Nordmark wrote:

>>>If IPv4 has such a notion from a congestion control
>>>perspective then something is broken.
>>>
>>
>>See RFC1112.
>>
> 
> RFC 1112 on Hosts Extensions for IP multicasting doesn't talk about
> congestion control.
> 
> Did you intend to type 1122?


Sorry - 1122.


> It does talk about both "local" and "congestion control" but the former
> is the subject of section 3 on IP and the latter is in section 4 on transports.
>  So I don't see where a notion of local is applied to congestion control.
> Do you have a reference to a specific section?


I'm referring to the use of broadcast on subnets.


>>>RFC 2002 (Mobile IP) uses IPinIP tunneling to make what you thought was 
>>>local (in the same subnet prefix) be capable of being anywhere in the Internet.
>>>
>>
>>Who said it had to be in the same subnet? IPinIP tunnels are 
>>point-to-point links, which means if they use subnets they are by 
>>definition misconfigured.
>>
> 
> I think this is an argument about a half-full vs. half-empty glass...
> 
> The use of local/remote in RFC 1122 section 3.3.1.1 is for the purpose
> of determine whether or not the packet should be sent to a router.
> Thus a possible interpretation of this is that "local" means exactly that -
> packets are not sent to a router.
> Another possible interpretation is that "local" means "close by" or
> "on the same high-performing instrastructure as the sender's network
> interface".
> Both are possible interpretations.
> Rubustness would lead me to pick a particular one of them in the context of
> congestion control.
> 
>   Erik
>