Hi Frederic, thanks again for your prompt and helpful reply.
Yes, but it could be tricky. bulkmarcimport.pl script invokes biblio.pm functions as you can imagine. But it do it correctly, ie it call several functions in order to add both biblio records AND their related item records. If you want to do it directly you have to hack bullmarcimport.pl script and understand how it works.
Then it is of course more straightforward to just produce unimarc records and feed them to bulkmarcimport... Thanks for having clarified this.
And also, are there functionalities in Koha's web interface that could be used to enrich an existing bibliographic record (or set of existing bibliographic records) with data fetched, say, from a Z39.50 server ?
(1) In Koha data entry page, you can perform a Z39.50 search. But you can't merge the result of a Z39.50 search with an existing biblio record. You can just overlap it or create a new record. (2) Another solution would be to use a standard z39.50 client in order create iso2709 file. Then you import this file into Koha with Tools > Stage MARC Records. In Manage Staged MARC Record, you can specify a matching rule. But here again, you can't choose to merge records (new and old). You just can replace, add or ignore. So there are two places in Koha where some improvements would help you to do what you need to do.
Well, not ure whether this could be automated. As I understand things, what I have to do is to first generate MARC records for all the titles we have in our catalog and then import them into koha. Then, as a second step, get the sbject headings for those titles from e.g. the national library and add them to the existing records. If anybody thinks this doesn't make sense, please do not hesitate to let me know. I'm completely new to the field so I ay very well be completely wrong. Thanks, Sébastien.