On 4/8/10 3:16 AM, Jukka Manner wrote:
[JM]: I personally don't like to idea that the DCCP header is changed when it goes through UDP encapsulation. Otherwise we are not talking anymore about just simply UDP encapsulation but rather about a whole new protocol. So the WG should either consider a) straight UDP encapsulation of DCCP (DCCP specific or generic? that is the question), or b) a new UDP-based DCCP-like protocol, as this draft proposes. But you shouldn't say that this draft is point a, which it is not.
I have to agree with this. I would rather see a straight UDP encapsulation with an API that enables it in the default DCCP implementation.
Even if it wastes bytes or duplicates fields, I would rather see as few changes to the DCCP packet as possible.
After this is specified, *then* you can start trying to combine fields to minimize packet size.
DCCP isn't deployed, period. I'd rather see deployment, first, and efficiency, second.
-a
