Re: git insists in using IPv6

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Sat, 18 Aug 2012 14:57:35 -0400
    From:        Julio Merino <julio%meroh.net@localhost>
    Message-ID:  
<CADyfeQVDNJZ1msJ_UttKwnmOF890H9Wnk_qSdS+HxqV76_jFgg%mail.gmail.com@localhost>

  | Anyway.  Taylor R. Campbell's reply, which just came in, holds the
  | real answer to my problems!

Yes, that was the same general kind of problem occurring in a more
complex fashion than I had assumed might have been likely.

This does point up a general problem we have with lots of apps though that
are failing to recognise what is the real problem with a connection
attempt, and mis-diagnosing the problem to the user.

No-one should have to resort to ktrace to fathom the cause of what should
be a simple error message - had git told you "certificate verification failed"
you would have fixed the problem in seconds I expect.

What does anyone think of the chances of a better error reporting API
(one that actually keeps track of the various connection problems and
reports the mmore likely to be the one that matters) actually being
accepted widely enough by the application authors (which means something
that would be acceted by all the *BSDs, glibc, and probably the more
important of the commercial *ix's (mac os & solaris I expect, perhaps
hp-ux)) that it would make sense doing the work to create it?

kre