Short version: The unicode for those modules is properly encoded in the OER2GO database, but there was a regression in the OER2GO backend code such that it doesn’t render unicode correctly in the frontend.
This is something that needs to be fixed on the World Possible side, but I know they’re very strapped for time at the moment. I’ll ping them again.
It looks like things have changed again, as the Unicode characters are showing up on OER2GO again. Could you check whether the issue you saw is still happening?
If it is, could you describe in more detail where the issue is showing up? You mention that XML parsing is broken. But that endpoint returns JSON. And the characters are all proper UTF-8. I’m able to parse it without issues in Python:
In [37]: print requests.get("http://dev.worldpossible.org/cgi/json_api_v1.pl").json()[138]["title"]
Khan Academy ( हिंदी, हिन्दी भाषा) - Kolibri
In [38]: print requests.get("http://dev.worldpossible.org/cgi/json_api_v1.pl").json()[149]["title"]
Sikana (Français) - Kolibri
Note that you can also load any Kolibri content channel directly from the primary source – Kolibri Studio, or via the CLI. We put a bunch of work into making them available as OER2GO modules as well, to facilitate preloading of specific content within the RACHEL ecosystem. But that’s not the primary source.