Ned Freed wrote:
> Repeating what I've said in previous messages:
(1) Simple mode canonicalization needs to ignore header folding. It also
probably should ignore header field name capitalization.
(2) Language needs to be added to the effect that noswp mode SHOULD be used
unless you know for sure that simple mode will work in your situation.
I also don't think interior white space removal is necessary in noswp mode, but
I can live with it being done.
If I can be allowed something of a reality check here, let's
be clear about one thing: the single hardest thing to achieve
with interoperability is getting the canonicalizations to agree.
Murray, Arvel's folks and I went through this twice as the
revs of the current dkim draft went by. Even fairly innocuous
changes required a fair amount of time to iron out, and in
fact the more innocuous the change the harder it was since the
diffs weren't always apparent. Take our current love triangle
and start adding dozens more sides -- which I suspect is happening
as we type -- and any changes become that much more difficult.
So the long and short of this is that I think that the bar ought
to be pretty high to change these. As in, there are fundamental
flaws.
FWIW, I'd say that we spent by far the most amount of time getting
"simple" to work. I don't know what Mark's experience was with
DK, but as I said, it's not always easy to judge how, um, simple
a change really is with canonicalization.
Mike