Re: Detailed approaches for demux

[Greg Daley <greg.daley@eng.monash.edu.au>]

Hi Francis,

Francis Dupont wrote:
>  In your previous mail you wrote:
>
>    When re-homing, also use the same flow label value with alternative 
>    locators.
>    
>    Demux at the receiver side is performed based on (flow label, src, dst)
>    
> => as the triple is chosen by the transmitter after a re-homing the
> receiver side should not rely on src (*) so there is a possible
> collision with another node using the same flow label value (**).

The original 3697 compliant flow label allocation is done on uniquely
on (label, src, dest), so it is easy to track these, on both the sender
and receiver, and they're unlikely to change.

This means that systems using sender side allocation should be able
to mark these as in-use.

Only if the receiver then allocates a label for multiple hosts (or
multiple conversations)(+), will there be a possibility of collision
with another (src, label) pair.

Regards,

Greg

PS +: for example, if there are > 2^20 conversations on the receiver.