I was wondering the same thing.
It looks like transifex supports regional subtags, but unless I missed seeing them, not script (eg. cyrillic / latin) or variant (eg. de-DE-1901) subtags.
http://www.rfc-editor.org/rfc/bcp/bcp47.txt
Maybe it’s only because nobody has done any yet?
according to the given rfc the following tags would be correct?
de-formal
de-informalMy question is does transifex support that?
We usually get rule information from unicode.org.
When we add support for a language we follow the BCP47 standard. The multiple language locales are based on region subtags.
Seeing “based on region subtags” doesn’t look hopeful (if it means only). But maybe the CLDR means there’s a chance.
Do you see “formal / informal” anywhere in that?
I don’t think so. I couldn’t find anything in the language settings of my Transifex test project that relates to language variations. This is all there is when I’m searching for German:
Yes, that’s what I see when I look around, some language-region translations but not any language-script etc. translations. (with maybe Chinese being an exception?)
@meglio
During your travels in web land have you found a way to specify cyrillic vs. latin?
Sorry I don’t understand the question. May you reformulate in other words please?
From what I see in transifex, language-region translations are supported. eg.
Portuguese (Portugal)
Portuguese (Brazil)
Spanish (Argentina)
Spanish (Latin America)
Spanish (Ecuador)
Spanish (Mexico)
Spanish (Spain)
But I don’t see anything like
Ukraine (Latin)
Ukraine (Cyrillic)
or
German (Formal)
German (Informal)
I noticed that your forum uses Cyrillic so I was wondering if you knew of a way to have a choice as to which variant to use.
Thanks for such a detailed explanation. Now I see what you mean 
The answer is NO - Ukrainian and Russian are single languages with no variations available there.
There are of course cases with huge regional differences, but those have never been formalized.
The number of percentage of world’s population speaking Ukrainian or Russian is declining, so I don’t think such a sectioning will ever happen.
Formal/informal variant is a big deal in languages with a T-V distinction (like German or Hungarian). Using a variant that mismatches with your audience can be alienating; even more so if the website that Discourse is attached to uses the other option. It seems in the years since this conversation has happened Discourse has switched from Transifex to Crowdin, so maybe this is easier to handle now?
Wrt language names, the standards-compatible approach would be the use of a BCP-47 private-use subtag, such as de_DE-x-formal. I see for German there is a plugin now, but that seems like a very hacky approach and I’d love to see a cleaner one (personally I’m interested in having an informal Hungarian variant).
Meanwhile, Discourse moved to Crowdin to handle interface language translations. I can’t find any reference to formal and informal variants but they do have regional variants. I wonder whether these regional variants are fixed by the system, or whether project admins can create new ones. If admins can, then it would be possible to create i.e. Hungarian (Informal). The software even can fill automatically non-translated strings between variants, so translators only have to focus on the sentences that differ with formal and informal.
If this process would technically work, then creation of formal/informal variants could follow the same process than the request for new languages.
As a speaker of various languages with formal and informal versions, I agree with everyone requesting this feature here (disclaimer: @Tgr and I come from the same project). Finding an English equivalent is hard, but imagine that the Discourse interface would be addressing everyone with the tone and the register of a servant in Queen Elizabet’s court. 
What is bothersome about the current “process” is that there is no common standard. Different languages have been translated to formal or informal depending on what the first translator(s) decided to do when they started translating.