I've done some major refactoring of Le Blanc, which is my pilot text for using <label> instead of <argument> for the marginal headings. That feature is working well, but I've had to do a huge amount of work to fix wildly proliferating <div>s in the original markup; they're still a bit of a mess, with chapter <div>s interspersed with other unidentified <div>s that have no apparent purpose, while some subsections in chapters are nested <div>s and others aren't. But a few more hours should sort it.

I've also generated a new schema, which includes <argument> in all the right places as before, but also includes <label> and adds the @type attribute to it (so we can distinguish <label>s used for marginal headings from those in e.g. TOCs.

In the process of this work, I discovered some aberrant behaviour in oXygen which threw me for a while. The old oasis-schema PI, linking to an RNG file, was not actually working, now that oXygen has switched to xml-model; instead, it was silently switching to validation with tei_all, which meant that many things would validate under the oasis-schema heading but fail to validate when xml-model was used.