[RFC] Cache DNS as long as the pycurl Transport is active

Wouter van Heyst larstiq at larstiq.dyndns.org
Wed Aug 30 22:56:40 BST 2006


On Wed, Aug 30, 2006 at 11:44:04AM +0200, Alex J. Champandard wrote:
> 
> John Arbash Meinel wrote:
> >it seems on someones machine [...] after the cache has expired, it seems 
> >to fail to connect.
> >  
> 
> That would be me! :-)
> 
> I'm on a standard XP setup without cygwin, using the latest Win32 binary
> of PyCurl (quite old).  I use bzrlib directly, but the problem is
> orthogonal to that... The process of updating multiple repositories
> takes over 60 seconds sometimes, and that reveals the problems.  Note
> that urllib is equally unreliable after a while on this same setup, so
> the urllib DNS cache hack helps.
> 
> I don't usually have any network connections or other problems, so I'm
> presuming it's either XP's handling of domain name lookups, or how
> bzr/python makes use of them on this platform...
> 
> 
> Has anyone else seen unreliable connections on XP machines?

Hmm, now that you mention it again, it could be the problems I had
branching bzr.dev on XP were of this same nature. Interesting, I'll have
to try again.

Wouter van Heyst




More information about the bazaar mailing list