Dear Pablo,

> Dear Christoph,
>
> just in case it might help to your meditation.
>
> First of all, lpaths are XPath implemented with Lua in ConTeXt.
>
> XPath seems to have as it primary purpose to address the nodes of XML trees.
>
> About your sources, the real issue here is to define whether you want
> TEI or TeX to be the format containing them.
>
> I mean, if you add explaining footnotes to your TeX file(s), XML won’t
> be source anymore.
>
> If you don’t want this to happen, you have to encode them in the TEI XML
> sources.

In my case, the connection to the original source will be lost. While we 
will still try to backport corrections, the connection will be lost at 
the very moment when we will annotate the source text.
 From this point of view, an independent source is created here anyway, 
so I have the freedom to choose between TEI or TeX.
At the moment I'm leaning strongly towards TEI, also because I could try 
to remove the annotations via XSLT once the work is complete, diff the 
result with the original source and improve it). And besides, 
XML+ConTeXt looks quite elegant from the distance of my ignorance. 
(Although reading xml-mkiv.pdf creates a pleasant frictional heat in my 
cerebral convolutions, but leaves at least as many knots in them as XML 
data have nodes).

> One last suggestion about proofreading. This is something I learnt from
> personal experience.
>
> The first reading is much better to catch errors in the text than the
> subsequent ones.
>
> Having the text properly formatted and printed on paper helps a lot to
> spot errors.

Same here: It's the paper!
(I for myself prefer for the first proof reading a page lots of white 
space, in monospaced font like Courier, while my proofreader wants to 
see the font that will be used in the end.)

Thank you very much for your suggestions,

Christoph