>To meet this use case, I want to extract all the headings indexed as
>dc.subject and display them at once (or in a paginated form).  I'm getting
>only the full dc.subject headings, not the separate terms that make up the
>headings.

Yep.

You should get whatever you would get if you did the equivalent search.

For example, scanning on 'dc.description any/stem a' will give you
stemmed keywords, whereas 'dc.identifier exact a' will give you exact
identifiers.

>The SRU scan operation requires me to submit a scanClause consisting of a
>complete CQL search clause:  index, relation, and searchTerm.

>I'm assuming that use of the "exact" relation in this context is intended
>to return full subject headings containing the searchTerm, so that
>"dc.subject exact durham" would return both "Durham (N.C.)" and
>"Architecture -- North Carolina -- Durham".  Please correct me if I've
>misread this point.

Exact is a complete string type of search.  If the database has
"Architecture -- North Carolina -- Durham" as a string which it thinks
is a subject, then yes it should be in there.

However scanning with a term doesn't return only terms that match, it
simply sets a location within the full list of terms to start at.

>But here's my real question:  is there no way within the SRW/U
>specification to extract all or an arbitrary portion of the dc.subject (or
>other) headings in a database without first sending a term to scan for?

The term in a scan clause is where to start from so that you can page
through all of the terms.  You can send an empty term, which should sort
to the top of the list, but scanning from 0 or ! is also a good bet.

There isn't a way to specify 'all terms' because for most indexes this
is a ridiculously long list.  You could however specify 10000000 terms
and see how many the server will actually return to you.  As a server
developer, you could simply not have a limit on your upper count of
terms that you'll return, and then query for some insanely large number.

Hope that helps :)

Rob