At 9:28 AM +0000 4/12/01, Charles Lindsey wrote:
However, all this ought really to have been spelled out explicitly in the
RFC, which certainly seems to be ambiguous as written, though your
interpretation would seem to be the only one which makes sense.
The canonical form of a media type should be described in the
registration for that type. This should include instruction on
whether text canonicalization should be performed. Most type
registrations are woefully inadequate in this regard, alas.
It's perfectly possible to have an application/* type that requires
newline canonicalization. It's possible to have a text/* type that
forbids it.
The only way to handle it correctly is to maintain a "table" of known
types and whether or not to apply text canonicalization on them, and
for types not in the table to apply heuristics such as Ned mentions.
Oh well.