RE: ACTION-502: Propose new model for RetrievalMethod in 2.0

Thomas Roessler wrote on 2010-02-09:
>>  That would be a circular reference in which you had a KeyInfoReference
>> as a child of the KeyInfo it pointed to.
> 
> So?

I didn't see any point in allowing a broken reference if it's easy to
preclude. Obviously other cycles and loops wouldn't be automatically
detected, just the simple case.

> My meta point here is that we should try to not constrain URIs in the
> Signature syntax unless we absolutely need to.

I prefer to allow only what's precisely needed and avoid the needless
complexity of the full syntax.

> If fragment identifiers in XML documents are using some kind of xpointer
at
> some point, then that should be fine.

Well, I already wrote 2.0 language explicitly precluding that for
References. Since this is mainly a 2.0 proposal, I was trying to be
consistent with that set of options.

> So, if we can get by by just saying "URI reference", I'd prefer that over
> defining our own subset of URI references that we permit.

XML Signature 1.x already defines that for References and it's quite a nasty
set of rules. I was trying to create a much more constrained subset that was
more or less the same as what's allowed for 2.0 selection/references. The
relative URI case was an oversight, I should have left that in.

-- Scott

Received on Tuesday, 9 February 2010 15:11:09 UTC