On 8/15/16 2:50 PM, Eric C Rosen wrote:
Thanks for your review.
On 8/9/2016 4:41 PM, Paul Kyzivat wrote:
it will be unlikely that a new implementer, schooled in the subject
matter, will be able to create a correct implementation.
The material in this draft describes running code. There are at least
two independent, interoperable implementations. The draft was written
after the implementations were done, and is a faithful description of
what was implemented.
While doing the implementation, it was discovered that the Ingress
Replication feature of RFCs 6513/6514 was under-specified in some
respects and unclearly specified in others. The purpose of this draft
is to document the procedures that had to be figured out during this
implementation effort.
I can assure you that this draft makes it a lot easier to implement this
feature!
Ah. Well, since it was written after the fact, by people who had
implemented without it, that doesn't constitute an existence proof that
a new independent implementation will be interoperable with the others.
(Before tweaking based on interop testing.)
But I do agree that the situation should be better *with* this document
than without it.
Of course, Ingress Replication is just part of a larger and more complex
system. The draft shows how IR fits into the larger system, but the
overall system remains complex.
Yeah. I am really glad that *you* are implementing it, and not *me*. :-)
I now have a greater appreciation for why internet-wide multicast is
generally thought not to work!
Thanks,
Paul
_______________________________________________
Gen-art mailing list
Gen-art@ietf.org
https://www.ietf.org/mailman/listinfo/gen-art
