specifying the <ds:X509SKI> element

I had a conversation with Peter Sylvester on the back channel about
the Subject Key Identifier certificate extension and its relation to
the <ds:X509SKI> element in HoK assertions.  I will paraphrase his
comments by proposing new normative language for the latter.  Instead
of what is now in the Holder-of-Key Assertion Profile, consider the
following:

"The <ds:X509Data> element MAY contain a <ds:X509SKI> element. If it
does, the <ds:X509SKI> element MUST contain a base64 encoding of the
DER-encoded Subject Key Identifier (SKI) extension of an X.509
certificate.  If the latter does not contain an SKI extension, the
<ds:X509Data> element MUST NOT contain a <ds:X509SKI> element."

Since the content of the SKI certificate extension (if it exists) is
not well-defined, the use of <ds:X509SKI> for the purposes of HoK
subject confirmation is more like <ds:X509SubjectName> or
<ds:X509IssuerSerial>, that is, it's only useful if there's an
underlying X.509-based PKI (which is out of scope).

I see two immediate advantages of this approach.  First, it simplifies
the normative language of the Holder-of-Key Assertion Profile, and
second, it aligns with the Metadata Interoperability Profile as it's
currently written.  The latter isn't a goal of the Holder-of-Key
Assertion Profile per se, but it's still an advantage of this new
approach, I think.

Thoughts?

Tom