Re: [AVTCORE] Comments on draft-ietf-avtcore-rfc5764-mux-fixes-03.txt

[Paul Kyzivat <[e-mail]>]

On 10/20/15 6:49 AM, Magnus Westerlund wrote:

> 3.  Section 1.4:
>
> I still don't see how what is written in this section results in any
> improvement. You authors have assumptions of how this needs to be
> implemented that isn't written down. Looking on this problem of
> demultiplexing I would build a table driven solution based on the
> ranges, where the packet is sent to the relevant protocol based on the
> table-lookup. If that protocol don't use that value, that will be the
> end of the road for that packet as invalid, rather than giving it to any
> other protocol. I get the impression that you want one to send the
> packet for testing by the protocols one after each other according to
> the order the protocols are listed. I still don't see that
> implementations will do that.

This has been my concern as well.

Look at it this way:

A natural implementation of this is via a switch statement. Each 
disjoint range of values is one case. (Detailed syntax varies by 
language.) A simple implementation of this may be implemented as a 
sequence of comparisons - a series of if/then/else such as you describe. 
a more time-optimal implementation uses a table lookup.

You shouldn't be mandating how this be implemented - only what the 
result is.

	Thanks,
	Paul