Scott Kitterman wrote:
FWIW, the original '/' problem came from a real (if complex)
record
Yes, I recall it, it was really invalid, but the problem wasn't
the slash, it was the invalid delimiter "@" in <macro-expand>.
who's to say what is weird.
If it's no letter, digit, hyphen, dot, or <macro-expand> then
it's weird. The usual dot between labels, not within a label.
Unfortunately that won't help you if you want to get it right.
I've been trying to feed the appropriate changes back into
pySPF and so it gets better as a result of this weirdness.
Does it offer real DNS queries ? Obviously it has q=txt for
redirect= and include:, but you need all relevant query types:
A, MX, PTR, SPF (TYPE99), and TXT.
So I need to make sure that the domain part has a dot in it.
Yes, or it ends with <macro-expand>, because then the required
dot could be hidden in the expansion. RfC 2821 rules, no dot
is bad. I've not checked whether John modified it in 2821bis,
but whatever he does for host = TLD won't affect v=spf1.
[check existence of MX etc.]
Exactly how to handle that aspect of it is one thing I am
still thinking about.
Maybe like redirect= and include:, but as warning instead of
error. And another warning if an mx results in more than ten
names - trying to catch cases where that's no problem because
all MX names are somehow covered by a CIDR is unnecessary. Bye