Re: Flow label versus Extension header - protocol itself

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 5-mei-2005, at 14:53, Francis Dupont wrote:


Francis, I get a headache from your weird quoting style.

I also don't get what the problem is. It's like we're talking about  
two different things.

I DON'T CARE WHAT A FLOW IS.

The only thing that's important is that everything with the same flow  
label has the same ULIDs and that there aren't overlapping flow  
labels between two hosts.

These requirements are easy to meet without undue inconvience for  
other flow label use.

The receiver doesn't get any say in the flow label value. Sure, it  
would be nice if the receiver could tell the sender which flow label  
it should use, but that's simply not implementable in a reasonable  
way, too bad, move on.

Anyway, in a post I don't think anybody read I suggested that we  
should use a sort of demux negotation. Since demux can happen on:

- unique locator sets
- transport protocol information (port numbers)
- flow label
- demux option in the packet

it should be non-fatal if one or more of the first three mechanims  
aren't supported by either the sender or the receiver since they can  
always fall back on another mechanism. So even if the flow label is  
used in a manner that is incompatible with the shim flow label use  
that wouldn't be a problem, the sender should just fall back on using  
the demux option if the locator sets aren't unique and port number  
demux can't be done.