Stripping comments

One task I have it to package our source XML files for use by integrators; one thing I'd like to do is first strip the comments from these files as they may contain sensitive information.

I was thinking that this could be done by processing each file through Saxon using a stylesheet which strips out comments and outputs the XML again. But rather than risk reinventing the wheel, I was wondering if anyone out there has implemented a DocBook comment stripper in their build process?

Thanks,

P.