Any reason www.python.org is slow?

Antoine Pitrou solipsis at pitrou.net
Sat Jul 10 10:47:09 EDT 2010


On Sat, 10 Jul 2010 16:28:19 +0200
"Martin v. Loewis" <martin at v.loewis.de> wrote:
> >> If your computer does have IPv6 connectivity, but it's broken
> >> (i.e. you have a gateway, but eventually packets are discarded),
> >> you see the IPv4 fallback after the IPv6 timeout. The IPv4 connection in
> >> itself then would be fast.
> > 
> > I think it's what most users experience when they are talking about
> > this problem. It manifests itself on many Linux setups.
> 
> This (*) is something I really cannot believe.
> (*) that there are many

Well, just take a look at the number of recipes for disabling IPv6
specifically in order to solve slowdown problems:
http://www.google.com/search?hl=fr&safe=off&q=linux+ipv6+disable+slowdowns&aq=f&aqi=&aql=&oq=&gs_rfai=

It is at least the third time that someone asks why python.org is
"slow", and that their problem is "solved" by disabling IPv6.

I disabled IPv6 myself, which solved similar slowdown issues.
The issues happened on *.python.org, *.google.com and a couple of other
domains; hence they weren't python.org-specific.
The issues happened with a Web browser but also with ssh; hence they
were neither application- nor protocol-specific.
The issues happened on two different machines, one hooked to a DSL
router, another with wireless connection to various outside networks.
Hence the issue is probably not tied to a particular hardware gateway.

I was quite surprised myself when I discovered that "solution". But it
really suppressed the frequent lag I had in some connection attempts
(including ssh connection to a rather close, mostly idle box).

It is possible that the way Linux (or some Linux setups: many of the
recipes above are for Ubuntu, I use Mandriva myself) handles IPv6
"connectivity" is suboptimal in some cases, and that connection
attempts don't fail immediately when they should. I don't have enough
knowledge to diagnose further.

Regards

Antoine.





More information about the Python-list mailing list