Dear Rick,
Thanks for your advice. I have checked: our MySQL Refbase database is encoded as "utf8-unicode_ci" and my Zotero export is also "Unicode UTF-8". Thus all is coherent and my previous errors are likely to be due to errors in encoding some initial references.
Sincerely yours,
    Daniel Thevenot

Le 18/06/2010 21:09, Richard Karnesky a écrit :
Accented letters are sometimes kept correctly,
sometimes transformed into several characters of no use.
Please refer to:

regarding character set issues.  Your MySQL character encoding must
match the encoding that you've configured refbase to use & these must
match the encoding of the file you upload.

We import such
references either manually or from Endnote database (where they are
correctly spelled), using Zotero on Firefox.
Zotero lets you specify your character encoding on export.  Please make
sure it matches the way youy have refbase configured.