on http://demo.refbase.net and http://beta.refbase.net, if there is an accented character in a field, the COinS produced by Refbase is not detected by Firefox OpenURL Referrer or by Zotero on my PC.
Note : to test just the COinS detection by Zotero without using the unAPI, try this link
Regarding the code, accented characters are encodes like this : %E9 for é.
On the website of my Research unit with a 0.9.5 version of refabse which is configured in french, COinS are correctly detected by Zotero or OpenURL referrer but accented character are encoded diffrently :
%C3%A9 for é for authors, %26eacute for é for the title.
Try this link : http://www.ceped.org/biblio/show.php?client=inc-refbase-1.0&wrapResults=0&author=d
Is urlencoding depending of a specific configuration of refbase and/or PHP ?
In Zotero's Forum, there is a post about character encoding in COinS : http://forums.zotero.org/discussion/933/character-encoding/
They speak about specifying encoding in COinS with ctx_enc=info:ofi/enc:UTF-8
ctx_enc is not present in refbase COinS.
I really don't know about the good things to do but it would be fine if accented characters in COinS produced by refbase were corectly interpreted by Zotero and other OpenURL plugins.
This has been fixed in the bleeding edge branch:
The COinS spec and the OpenURL spec do not make UTF-8 mandatory, although UTF-8 encodings are suggested in the COinS guide: