Result of artifact type discussion

[Scott Cantor <cantor.2@osu.edu>]

The focus call participants discussed the various historical artifact types
across SAML 1.x and ID-FF 1.2 and attempted to determine the optimum
approach for 2.0.

My take on the sense of the group was:

- having a fixed length option is useful and important
- fewer formats are best
- metadata needs to be reviewed and possible amended to address whatever
type is chosen

Because the ID-FF artifact type is essentially like the original fixed
length SAML artifact type with the addition of a predictable mapping from
providerId -> SourceID, that seems like the best choice to use for 2.0.

Formally, we would define a new type with a new URI that replaces all three
existing types, that maintains the ID-FF presumption that the SourceID can
be generated by hashing an artifact producer's unique identifier. I would
also propose that we add the hashed value to the provider's metadata.

This solution has the advantage of requiring little or no change to most
aspects of the mapping process for both ID-FF and SAML 1 implementations,
since SAML 1 just assumes the SourceID is arbitrary anyway.

Barring objections, we can publish a new bindings draft with this defined.

-- Scott