On Tue, Jun 27, 2006 at 09:57:04AM -0000, John Levine allegedly wrote:
I'm with Paul, the more marginal features we can remove, the more
likely people are to implement our spec correctly.
I too am a big fan of removing marginal functionality from the core.
Given that over the last year or so no compelling cases have emerged
for *this* particular canonicalization (eg, discovery of a widely
deployed munger), it's not even clear that if we do later find a need
for another body canonicalization, that "relaxed" will satisfy that
need.
In other words, we are now in the mode of guessing that some future
need will be serendipitously solved by this functionality. Could it
not as easily be the case that this future problem needs a
canonicalization other than "relaxed"?
As I recall, all of the canonicalizations were based on conjecture
about what happens to mail in transit. Certainly conjecture by
experienced folk, nonetheless, our question now should be: was that
conjecture correct?
One other point. If we assume wide-spread adoption, the requirement
for canonicalization should diminish over time as MTA programmers
internalize that mail munging is bad. Conversely, if there is no
wide-spread adoption, we're unlikely to discover this future need.
Mark.
_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html