Schematron

Over the last week or two the activity on the Schematron list that Rick Jelliffe maintains has risen quite sharply. Rick made a statement that he was back on the case and shortly thereafter posted a beta implementation of ISO/IEC 19757-3 using XSLT.

I've always admired Schematron. It has the simple ideas (with a complex implementation) that attracts me. It's uses are currently being collated by Roger on the xml-dev list. It seems to be capable of reaching the corners of the validation scene that other tools either don't, or can't approach. Bottom line, it is useful. Very useful if you care about the nuances of validation. Or perhaps if you have a pointy headed boss telling you that we (meaning you) simply must check that this (whatever this is) must never happen again. I'm guessing at this scenario of course. Business rules are a natural for Schematron. I'm hoping to scrape up Rogers findings and add them to my website when he's ready. With his permission of course.

Not being much help with the XSLT (it's quite devious to say the least, and Rick is coping well), I looked for documentation and found none. As with XSL-FO, I thought I could help, so I've started to put a user guide together. Initially aimed at an XSLT 2.0 implementation, I may add something on Uches Python implementation if I can get my head round it. I'm unsure if he's developed it further than his initial work. Ken Holman has put together an XSLT implementation for UBL, apparently with one or two tweaks that are required for that market, hence less likely to be of general use. One such I believe is to terminate on the first error. We'll see how that one develops. My goal was something that could be extended by users, using XSLT, yet without such extensions was as near the standard as possible. That seemed (to me), about maximally useful.

I'm enjoying writing again, especially since it is work I'm investigating as a new user. I've not touched Schematron since Rick first announced it last century. Then it made my head hurt! I liked it, but its use of metastylesheets (as Mike Kay has labelled them) was then quite new to me. I could see the potential though, even then. With the additions of the last few years I'm confident Schematron can move on significantly. I just hope my tiny contribution helps.

The only bit I'm not satisifed with is the Annex D, Schematron Validation Report Language, SVRL. Seems this was wanted by others on the WG, less so by Rick. The result is an informative annex, providing a half hearted XML output language. I like the idea, but I think it wants matching better to both Ricks implementation and candidate output formats. I may well try to help here. I'm sure my submissions will be scrutinised on the list! There are some well informed people listening in.