Fallback to cs

1. Is this even possible in MediaWiki? I always thought about fallbacks as a chain, not a (possibly cyclic) graph.
2. I guess it would be quite surprising for Czech users to see messages in Slovak, as this is something no other software does, I’d say. And, given both languages have quite a good coverage, I believe, it is not that necessary. OTOH, in cases where an untranslated message in English would appear, having a message in Slovak would probably be… more useful for some users? (Even though surprising, as I said.)

So, I am not sure; it’d probably not hurt a lot, at least.

Could we have a rough count of messages translated in Slovak but not in Czech (or vice versa), just for illustration?

1. Yes, it's possible. We have for instance a reciprocal fallback between pt and pt-br.
2. Coverage is ok for the most frequent extensions, but not for all of them. Maybe this change is going to affect/help mostly non-Wikimedia users. Is English less surprising/annoying, for the average user? (I'm not aware of data other than language proficiency statistics, to answer this question; maybe we can think of something else too?)

Special:LanguageStats/cs vs. Special:LanguageStats/sk can give an idea: MediaWiki extensions are respectively 54 % and 27 % translated, so we have at least 27 % messages in sk which would use the fallback. I'll try to make a list to see how much overlap there is.

@Mormegil: It is better than nothing. Czech users are still more like to see Slovak translation if Czech translation is missing and vice versa than to see English/German/Polish strings or nothing at all.

I made a list of messages: all cs, all sk, difference (comm -3). About a thousand messages are available in sk but not in cs, and about 7 thousands in cs but not sk.

Thanks, but it seems not to match reality in a few random tests I used. E.g. Notification-add-comment-yours2 is listed in sk and not cs, while it exists in both (and is an outdated/unused message anyway). Dtto Centralauth-editgroup-subtitle or tpt-desc. OK, User-profile-preferences-emails is in sk but not in cs, but it is unused anyway. Dtto ga-threshold. Whoa! Finally I have found Security-restricttogroup which seems to be a correct example. But the overall validity seems to be not very good by this random test.

Anyway; as I have said, I am not against the idea, just that I do not consider it that much important or useful. If someone does, by all means, go ahead.