> If the choices to extend an existing protocol are between adding P-bit
> and assigning a UDP port per each protocol, I still think the P-bit
> would be better since it is the more generic solution, but is going to
> be more work to ensure backwards compatibility if flags can be ignored
> on RX.

But we are not starting over from scratch. There are implementations that have 
been around for years. We don't have to luxury you say. And as I said before, 
every time you add a feature you will need to change the implementation anyways 
and deal with compatibility. So making a header general with tons of TLV 
options isn't necessarily the right approach FOR A DATA-PLANE. Versus just 
changing the header for the new requirements you see in front of you at the 
time you design the header.

And I mentioned this before at a previous NVO3 meeting. In ~30 years since the 
IP header was defined, how much have we, in practice, used IP options.

Dino

_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to