>>>>> "NB" == Norbert Bollow <firstname.lastname@example.org> writes:
NB> As I said, I'm going to do the Bounce-handler.
Ah, I had forgotten. You had gone off to do your thesis and I assumed that
would keep you away for quite some time.
NB> I'll look at the next snapshot which you put out, assume that any
NB> aspect of corner-case RFC compliance which is not in the snap-shot is
NB> something which you'll not have time to do, and try to document those
NB> instances which I'll notice.
OK. I hope to have most of them cleared up, or at least tightened up.
(I'm sure there are illegal addresses that are accepted by the validator,
but these should be very contrived cases.)
Now to tackle the problem of MIME-tools not 2047-encoding headers. Right
now it (MIME-tools and hence Majordomo) follows the GIGO rule; if 8-bit
headers come in, they go out. This is suboptimal, but it's also impossible
to rectify in a sane manner. Why? Because 2047-encoding requires that you
know the character set. You can have multiple character sets in one
header. The vast majority of sites will do the right thing with respect to
8-bit headers, so if we choose to encode them in some random character set
we run the risk of mangling a header that is perfectly acceptable to most
I'll probably do the mangling (by running over the headers and calling
MIME::Words::encode_mimewords on each) based on some per-list configuration
Note that if 2047-encoded headers come in but 8-bit headers go out, that
means I screwed up.