Mats Wichmann writes:

...

...

...

One possibility is that the Linux getaddrinfo() is thread-safe, but only by way of a lock that only allows one request to be outstanding at a time.

The next step should be to get the getaddrinfo() source code from glibc and see what it does. It's open source, hey. :-)

I can dig around a bit, but I have to figure out what I'm looking for.

[MvL]

I think that part is already settled: getaddrinfo, on Linux, is thread-safe.

...

On the failure platform, are we sure Python is using the native getaddrinfo, not the Python-supplied one?

Correct.

I think the remaining question is: Even if the GIL is released around getaddrinfo - why is the performance of Jeremy's test script still that bad?

I tried to read the glibc getaddrinfo() source, but it looks like it would be a term project... It could be that it's just doing a lot more interaction with a DNS server. I believe that Jeremy suspects that the test program isn't just slow, but that one slow thread actually blocks all other threads from making progress. If that's the case (we don't know for sure), we're looking for a bottleneck in the getaddrinfo() code that somehow holds a resource needed by all threads calling getaddrinfo(). --Guido van Rossum (home page: http://www.python.org/~guido/)