On Aug 11, 2017, at 9:27 AM, Juliusz Chroboczek <[email protected]> wrote:
> Can we please agree that this document has no business mandating 
> round-robining?

The point of the text on round-robining is to avoid a situation where one 
provider's answers wind up being preferred over another provider's because of a 
difference in the number of recursive resolvers they offer.   I wrote it the 
way I did because the mental model for how to do it that sprang into my mind 
when I was writing it was the round robin model.

I don't have a strong preference for that model, although I think in practice 
it works, particularly if you take a happy eyeballs approach to it rather than 
timing out before switching.   That is, you don't ask all resolvers at 
_exactly_ the same time: you ask one first, and give it enough time that if 
it's operating in spec, you will get an answer before you ask the next one, but 
you ask the next one long before the query to the first one has timed out.

The bottom line is that this is an implementation detail, and I think you are 
right that it should not be specified the way it currently is.

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

Reply via email to