As more and more digital texts get produced and then repurposed 
(curated?) into TEI, I am sure more and more people have had to face the 
problem of how properly to represent their pedigree (the text's, not the 
people's) in the sourceDesc. But I haven't yet found any very clear 
indication of recommended practice in this respect, or not one I like 
much at any rate.

Here's a far from unusual scenario. The project is producing a 
collection of literary texts in TEI, many of which are already digitized 
in page image, or HTML, or some other non TEI format. They may even be 
in TEI, but it's not the same as the TEI we want in our project. The 
project has defined a rather strict and specific TEI schema, and 
everything has to be converted to it. Consequently, it needs to record 
in the sourceDesc up to three bibliographic descriptions -- one for the 
digital source used, one for the print source from which that digital 
version derives, and possibly one or more others for sources used to 
modify the primary digital source.  I don't think it's good enough just 
to list the three bibls (if bibls they be) because that loses 
information about the relationships amongst them. So here is an example 
of how I am thinking of doing this:

> |<sourceDesc> <bibl> <ref 
> target=""> Tatiana Leïlof 
> roman parisien (édition numerisée) </ref> <publisher> / 
> Bibliothèque nationale de France </publisher> <idno 
> type="ARK">12148/bpt6k931128v</idno> <relatedItem type="printSource"> 
> <bibl><title>Tatiana Leïlof , roman parisien, par Édouard Rod</title> 
> <publisher>E. Plon, Nourrit et Cie</publisher> 
> <pubPlace>Paris</pubPlace> <date>1886</date> </bibl> </relatedItem> 
> </bibl> </sourceDesc> |

I think this shows rather nicely that the source of the text in the 
header of which this appears is a digital text published by the BNF with 
the identifier shown, the print source for which is the title published 
in Paris in 1886.`Now suppose that the digital source used for the 
project has been collated with a (fictitious) 20th c edition to create 
our new TEI version. I can just add another relatedItem within the outer 
bibl, distinguishing it by means of its  @type attribute:

|<relatedItem type="collatedWith"> <bibl><title>Tatiana Leïlof , roman 
parisien, par Édouard Rod</title> <publisher>Gallimard</publisher> 
<pubPlace>Paris</pubPlace> <date>1902</date> </bibl> </relatedItem> My 
question to the list is : does this look reasonable? and if you were (or 
have been) faced with this scenario, how would you deal with it? I know, 
I know, you'd use RDF. But say you want to humour an old man, and do 
everythng in TEI :-) |