I'm still puzzled by this diagram. Your description of a "customized subset" suggests that it may validate documents which are not valid against tei_all, whereas the blob for it falls squarely within the TEI subset box. As far as I'm concerned, I would make it even simpler: any customization that validates files which are not valid against tei_all is an extended subset. It extends the TEI by providing options (perhaps new elements or attributes, perhaps just new content models for existing elements) which were not available before. I think the definition of TEI conformance should be that all files valid against a schema generated from the customization also validate against a tei_all schema generated from the same P5 subset used to create the first schema (so the TEI versions used must be the same). Anything else is an extension. This is a purely mechanical test, of course. It doesn't check whether you're using <title> to tag measurements or <name> to tag bold text. I think adherence to the spirit of the prose definitions and descriptions needs a different word to describe it (and a human to judge it). Cheeers, Martin On 2017-03-17 06:24 AM, Lou Burnard wrote:
I've been thinking about what the Guidelines say about conformance in chapter 23, following Michael's spate of tickets and the subsequent debate last month. It seemed to me it might be helpful to establish whether the Council agrees about what the notion of conformance *ought to mean* before trying to make sure that the text of the Guidelines express it. So I have prepared a little (really little!) document for you to read and disagree or (hopefully) not with. All comments welcomed.
The document is at http://lb42.github.io/W/conformance.html (there was an earlier version on my foxglove blog, but now that I've got my ceteicean foo back I'll be maintaining this document on github instead)