On Jun 3, 2011, at 3:58:26 PM PDT, Daniel Stenberg wrote:
> It is somewhat implied in your mail, but let me ask you anyway: have you experienced any downsides with this change?
We have not yet deployed this change (we reverted to c-ares 1.6.0 in our currently deployed software to avoid the problem), but I have not encountered any problems in my testing with this patch. The one downside is that, because so few domains have actually obtained an IPv6 address, it often requires two DNS queries to resolve them.
That is the reason for my more extensive patch that I mentioned in my later email, which tries IPv4 first and only tries IPv6 if that fails. But that later patch has had only minimal testing. And, as Uli points out, that is a simplistic approach which does not take into account all the nuances, such as the ones he references in RFC 3484. But I suppose one might argue that the current c-ares implementation is also a simplistic approach.
Cheers,
--Charlie
-- Charlie Fenton charlief_at_ssl.berkeley.edu BOINC / SETI_at_home Macintosh & Windows Programmer Space Sciences Laboratory UC BerkeleyReceived on 2011-06-04