Let me try to describe the situation I have in mind fairly clearly.  I have
a relational database.  It has a table of "access points".  This table is
the "index" to the bibliographic records.  It is also the "index" to the
authority records.  The rows in the table tell you what the search terms
are in my database.  They also tell you (through a bit of XML magic) when a
search term is a reference to another search term.  So, when someone does a
scan, I return data from this table. And only from this table.  I think it
is helpful to the user to include the reference information that I already
have in the table.  It is no effort for me to do so; in fact it is
significant effort for me to exclude the reference information.

I don't want scan to preclude this behaviour.  And the single "alternative
term" element doesn't have rich-enough semantics to let me express the
variety of references I already have.

I can also, if I want, use this table to collapse and expand heading
hierarchies because each row has a hierarchical level indicator and a
pointer to it's parent if it's not a root.  So I can easily use this table
to do simple thesaural navigation.  I don't have a separate thesaurus
database.  I don't have a strong requirement to do this right now, but I
can see how it will be useful in the future.

If I wanted to refer to a separate database that could be used to provide
thesaural access or a finding aid to the data in this database, then I
agree, this would be outside the scope of scan.

j.