>> Notice the addresses don't match the configuration: 212.254.26.75 is
>> an alias now, and 212.254.26.66 is not an alias anymore!
> Why do they not match the configuration?

That's the question, yes.

> The kernel doesn't really have a concept of "aliases", just multiple
> addresses on an interface.

But the first one on the list *is* special in a few ways (for example,
it's the one that gets replaced if you ifconfig an address on the
interface without saying "alias", and it's the one that gets used as
the source address of source-not-specified traffic going out that
interface).

> I'm not sure why the order changed, but a program which depends on
> the order is clearly buggy.

Yes and no.  A program which depends on the order *per se* might be
buggy (though I could argue against that).  A program which depends on
other things which the order affects, such as the two properties I
mentioned parenthetically above, I most definitely disagree - I don't
see it as buggy for a program to, for example, expect to get the
primary address as the source address on connections made without
binding a source address.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B