> <a href=3D"http://www.python.org/ ","http://www.python.mirror1.org ","http://=
> www.python.mirror2.org/">Python Powered</a>

This is not well formed XML and lexically invalid SGML.  Accounting for that,
what you have here seems to me to be a single compound resource locator, not
three separate ones.  As its a resource locator, it is not a W3C issue,
but rather an IETF one.

To be honest, I find it difficult to think of many cases were this sort of
mechanism wouldn't be an abuse of the internet (if the site is popular,
caching three copies at your ISP is a waste of their resources; if
it isn't popular, it is likely that the bottleneck will be the client
in almost all cases).

However, the right way of doing this is probably to use rel attributes.
Typically, where there are mirrors, the initial link is to the mirrors
page, so that needs one type of relationship, and the mirrors pages links
to the individual mirrors, which would use a different relationship type,
to indicate that all URLs with that relationship on the page represented
the same resource.