If you look at other localization files, no non-ascii chars are HTML encoded. This is the target of using UTF-8 to keep all chars in their raw format. Why should we make an exception for German? And even though, there's no significant advantage of having umlauts HTML encoded. Since RoundCube also uses localized strings in Javascript alerts and plaintext output, HTML encoding would mess up these.
Okay then. I wasn't really thinking globally there, but if stuff is handled the UTF-8 way, then we'll hopefully be okay :)
I'm Swiss and it's my personal wish to have a Swiss German translation. Don't worry, I'll maintain these files myself and nobody else will have any extra work with them. And what are the "inconsistencies" you're talking about?
Roundcube is your baby, so do as you please here :) I was looking at both tranaltions in translator.roundcube.net and I saw that de_CH is missing 'deletemessagesconfirm' and that some translations are different although they don't really need be IMHO, like 'loading' is "Daten werden geladen..." in de_DE and "Lade..." in de_CH.
Which reminds me of another issue: Language hacks such as "Select message(s)" for both singular and plural forms don't work in most languages. I think having separate labels such as 'deletemessage' and 'deletemessages' is not an elegant solution, but maybe we could make the labels part of a two-element array, selecting element 1 for 1 item and element 2 for more than one item.
Cheers, ~Mik