On 2005-07-20 08:45:43 -0700, Dave Crocker wrote:
A premise to the wsp optimizations is that they do not affect the
structure of the message. So canonicalization of wsp is only for
that has no structural impact. For example, wsp at the beginning
of a line in the rfc2822 header has very big structural impact
because it means the line is a continuation of the previous
line's header field.
In other words, the reference "line folding" for canonicalization
is really a rule about structure-related processing, as distinct
from "data-related" processing.
I think think means that the manipulation you cite with regards
to MIME delimiters is to require canonicalization to be aware of
MIME structure constructs and to make sure that they are
preserved?
Making sure that they cannot be changed without making a
corresponding change to the canonicalization is the important bit --
this can be done without being particularly aware of them, as is
demonstrated by the simple canonicalization.
That said, I wonder how relevant the kinds of changes that nowsp is
optimized for really are, e.g., re-flowing of e-mail content. In
particular when one compares them to the effects of transformations
such as the in-transfer re-encoding of MIME bodies between, say,
8bit and quoted-printable.
Regards,
--
Thomas Roessler, W3C <tlr(_at_)w3(_dot_)org>