Received LR's latest NLM2TEI conversion XSL, and tested it with my NLM 2.3 files. There are about four problems in the resulting output, all minor, and all fixable. I sent a detailed report back to him. Some of them are actually due to restrictions in my own teiJournal schema which don't necessarily need to apply to TEI output from his XSLT, but others are due to more general issues. We're close to having working round-trip code now. It's partly an academic exercise, but it does also make a good testing procedure for the TEI-to-NLM code which teiJournal must have, and for his NLM-to-TEI code, which has general value.