On Tue, Apr 11, 2017 at 09:31:41AM -0800, Melinda Shore wrote:
On 4/11/17 9:18 AM, Nico Williams wrote:
One could give a lot of advice for design of protocols with
"friendly" middle boxes. Merely saying "hey, they are good" is not
enough. We might want to revisit end-to-end protocol design as well
(e.g., maybe ICMP isn't working so well; what can we do?).
There have been a number of efforts to provide mechanisms for
applications to communicate explicitly with middleboxes. None
has gotten any traction, and for the moment it looks like
anything that requires changes to middleboxes along those
lines is unlikely to be successful. That said:
Sure, but if we wanted to have an Informational RFC describing all the
goodness and badness and history of middle-box-aware protocools, that
might not be bad, and we could detail all those protocols that failed to
get traction and why. I think such a document would end up favoring
end-to-end protocols, even if it didn't start out with that as a goal :)
IMO the IETF must not publish draft-dolson-plus-middlebox-benefits as
it is today.
No, clearly not. I'm actually not sure I see a lot of benefit
to publishing a more balanced document, either, in the sense that
it's not likely to lead anybody to do anything differently.
I'm not advocating for publishing a more balanced document. I only
advocate not publishing draft-dolson-plus-middlebox-benefits. If the
authors come back with a more balanced document, I might be willing to
support it, though I can say right now that I would expect the result to
favor the end-to-end model, and if it ended up advocating middle boxes
much more intelligent than routers then I'd be surprised.
Nico
--