ietf
[Top] [All Lists]

Re: I-D Action:draft-rosenberg-internet-waist-hourglass-00.txt]

2008-02-13 14:50:15
I would disagree about "no drawback."
The constraints are:
don't invent new transports
have congestion control

and, for an important part of the traffic, do not have retransmission or 
significant transmission delay (i.e. perform congestion control by 
dropping packets.)

Now, we have SCTP modes and DCCP which can address those needs over IP.
We do not have a defined protocol that runs over UDP that meets those 
constraints.
We could just use UDP.  But the congestion sensitivity aspect was 
actually an important part of the work.
And the scoping (very near, near, or anywhere) was defined in the 
negotiation with the IESG to give us bounds on the work and a framework 
for making these sorts of decisions.

So ignoring the bounds would be quite inappropriate.  And the problem is 
not solvable within the waist you have defined.
Which leaves us with either getting the work done, using IP and suitable 
transports, or extending a very long process much longer.

Not a good choice.

Yours,
Joel

Jonathan Rosenberg wrote:
Well, if history is any guide, eventually people will in fact want to 
run this from someplace a little farther away, and then you're in big 
trouble. So, I think the advice remains the same. There is no drawback 
to having it over UDP to start with - it works when there are no NAT, 
and it can work when there are NAT.

-Jonathan R.

Joel M. Halpern wrote:
However, I would really like to reinforce the point from another note. 
There are quite a few contexts where the ability to run a sensible 
transport directly over IP is indeed very useful.  For example, the 
ForCES working group scope is limited (by chart) to the case where the 
control element is near the forwarding element.  I am not worried 
about there being a NAT between those.  So SCTP or DCCP over IP is 
very relevant.

Yours,
Joel M. Halpern


Jonathan Rosenberg wrote:
I wrote this because of a discussion that happened during behave at 
the last IETF meeting in Vancouver. There was a presentation in the 
behave working group on NAT ALG for SCTP - when run natively over IP 
- and I found the entire conversation surreal. The entire problem 
would have been moot if SCTP had been designed to run over UDP and 
not IP.

So apparently its not obvious to everyone that you cannot design 
protocols natively ontop of IP.

-Jonathan R.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
http://www.ietf.org/mailman/listinfo/ietf


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
http://www.ietf.org/mailman/listinfo/ietf

<Prev in Thread] Current Thread [Next in Thread>