I have been in long pursuit of an XML-based query-able data store, and despite continued searches and evaluations, I have yet to find a solution that meets the my needs, which include:

1. Data is wholly contained within XML nodes, in flat text files.
2. There is a "native" - or at least unobtrusive - method with which to perform Create/Read/Update/Delete (CRUD) operations onto the "schema". I would consider access via http, XHR, javascript, PHP, BASH, or PERL to be unobtrusive, dependent on the complexity of the set of dependencies.
3. Server-side file-system reads and writes.
4. A client-side interface element, accessible in any browser without a plug-in.

Some extra, preferred (but optional) requirements include:

1. Respond to simple SQL, or similarly syntax queries.
2. Serve the data on a bare bones https server, with no "extra stuff", either via XMLHTTPRequest, HTTP proper, or JSON.

A few thoughts:

What I'm looking for may be possible via some Java server implementations, but for the sake of this question, please do not suggest that - unless it meets ALL the requirements. Java, especially on the client-side is not really an option, nor is it appealing from a development viewpoint.*

I know walking the filesystem is a stretch, and I've heard it's possible with XPATH or XSLT, but as far as I know, that's not ready for primetime, nor even yet a recommendation. However the ability to recursively traverse the filesystem is needed for such a system to be of useful facility.

At this point, I have basically implemented what I described via, of all things, CGI and Bash, but there has to be an easier way. Thoughts?