Re: [v6ops] An Update to Happy Eyeballs

This is also a area where fastest is the enemy of faster.

There is zero reason for making async DNS a MUST.  A MAY would be
fine for the DNS lookups.  AAAA lookups vary rarely fail to resolve
these days which was the original reason for splitting the two
lookups.  It also adds a lot of complication for very little gain.
It also doesn't preserve a preference to IPv6 as it depends on cache
content not transport reachability, 50ms is not enough time to take
cache contents out of the equation if A is cached and AAAA isn't.

nodes are supposed to impliment destination address selection policy
rules and that includes deciding whether to talk IPv4 or IPv6 first.
This document breaks that for no good reason.

Mark

In message <DE424897-E36A-4441-8F26-009AB0AD094E@apple.com>, David Schinazi wri
tes:
> Phillip,
> 
> By design, Happy Eyeballs cannot be implemented using blocking
> networking calls in a single-threaded C program, as it involves
> racing parallel operations. DNS is no different than TCP in this regard.
> 
> If an implementation wants to be compliant with this specification,
> they'll need to support asynchronous networking. If an implementation
> has limitations that prevent it from being compliant, that is fine,
> you don't need to support all RFCs to connect your code to the Internet.
> 
> If there is consensus in the working group to relax the requirements
> language in this document, we're absolutely open to that. But I don't
> believe we should restrict ourselves to getaddrinfo() or any API for
> that matter.
> 
> David
> 
> 
> > On Mar 13, 2017, at 05:49, Philip Homburg <pch-v6ops-6@u-1.phicoh.com> wrot
> e:
> > 
> >> 
> >>   We've incorporated them into -01, please let us know if they
> >>   were properly addressed.
> > 
> > In my opinion, there is no justification for declaring a single threaded
> > C application that uses getaddrinfo non-compliant.
> > 
> > So this draft doesn't address my concerns. Just calling getaddrinfo should
> > be enough. 
> > 
> 
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka@isc.org