On Thu, 9 Feb 2006, Shmulik Regev wrote:

> From what I can tell, through the use of persistent connections the wrong
> DNS entry is (a) saved; and (b) reused with every new request. If the
> implementation would prune the cache after an unsuccessful request, i.e. one
> that failed prematurely my problems would be automatically solved. Do you
> think this should be the behavior of libcurl?

Yes!

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html