saml-metadata draft comments

I have a few comments reguarding the metadata spec, to whit:

- was the ommition of the BNF of entityIDType deliberate?

- i noticed you dropped the cacheDuration attribute.  when the
resolution protocols are re-introduced, i think we will need to include
this (which is a more common caching model in resolvers, AFAICT)

- 2.1.5.8 ds:sig indicates pubkeys for signing are exchanged outofband
with a non-normative must.  i think this statement is unneccesary,
instead "keys required to verify the signature is out of scope for this
specification"

- SPDescriptor drops support for the pulication of multiple
assertionconsumerserviceurl. the uses case (from RSA) which supported
this design choice, i think, are still quite valid, imho

--- peterd