> if you are proposing that the IETF should investigate ATP, have you
> submitted the proposal as an internet-draft?
No.
> If not, why not?
I myself have some unsureness on ATP.
1. There're too many contraints in the tranditional TCP/IPv4 internet environment. So
ATP is going to be optimized for IPv6. Yet there're quite a lot debates on IPv6
itself. For example on the flow label field.
I like IPv6, Of course.
2. I think it may give the upper layer application users or programmers great
convenience if key features of RSVP, ISAKMP, IPSec and IPComp are consolidated in a
single set of APIs. The APIs actually represent the services provided by the transport
layer.
But maybe it's more convenient to devise a session layer protocol. I want to raise the
consideration.
3. ATP fundamentally changes the meaning of the diffserv/traffic-class field in the
IPv4/IPv6 header.
There's a saying that IP will definitely fail if it tries to provide as complex QoS
features as ATM CBR, ABR, VBR or UBR. I'm afraid I agree with the opinion. So I assume
it is feasible to provide only two classes of service: real time, and best effort. The
assumption itself may raise heating debate.
I'd like to defense the idea of ATP and make necessary update before it becomes a
proposal (or does not become a proposal).
