I am now (fingers crossed) at the end of three days of XSLT which was quite unnecessary. I've been blogging for about five years now, and have tweaked the software that generates the blog more than once. I came across an irregularity and decided to correct it. Easy. Identity transform, one template and ...screw up. My identity transform didn't include attributes. Grrr. I was using the collection() function to iterate through all the xml files. I had a backup so thought I could search out all the elements in error and correct them. Then I found more 'changes' over time. This meant that rather than assuming two maching files (in terms of element count) I had a mismatch. That meant manually adjusting about 100 plus files. Slowly I stopped 'restoring' the corrected files and built more backups. I basically didn't trust anything. Finally finished this morning.

Lessons learned? Import a known good identity transform. Don't try and knock it off in a minute whilst thinking of other things.

Run a thorough Schematron type of check for elements which count. Validation takes on a whole new meaning when used like this. I guess I'm slowly catching up with Rick Jelliffe in terms of my understnading of what (can be) needed from validation.