--On July 1, 2005 4:44:23 PM +0900 Martin Duerst <[EMAIL PROTECTED]> wrote: > > The reason for this is to make sure we have interoperability > with a mandatory-to-implement (and default-to-use) canonicalization, > but that we don't disallow other canonicalizations that for one > or the other as of now not yet clear reason may be preferable in > some cases in the future (but in your wording would prohibit > the result to be called Atom at all).
A potential future reason that we can't even characterize isn't enough reason for me to support this. If we discover weaknesses in the canonicalization, we'll need to change Atom anyway. Explicitly making room for future incompatible canonicalizations doesn't make any sense to me. What is the point of calling something "Atom" when it uses a canonicalization which prevents interop with legal Atom implementations? wunder -- Walter Underwood Principal Architect, Verity
