One issue I already discussed in private with Mitko is that all names on the OSM map of Bulgaria should be both in Cyrillic and Latin script, see https://wiki.openstreetmap.org/wiki/Multilingual_names#Bulgaria For each name in Cyrillic, we should also add one in Latin script for map users who can’t read Cyrillic (a good renderer will display a name in Latins script if the user has set a language preference for one that uses Latin script). Of course for any name in Latin script, we should have one in Cyrillic too.
For geographic names, names of historical persons and cultural items, we have to use the rules of the transliteration law.
For unofficial names, such as for shops, companies, etc. we should use the “on the ground” rule: the name that a shop displays or a company uses should go into the name=* tag: if it’s shown in Cyrillic, it should be tagged in Cyrillic, if it’s in Latin script, it should be in Latin script in the name=* tag too.
If the name used is in Cyrillic, it should be added in Latin script as well using the int_name=* tag. I think we should again use the “on the ground rule” and spell it like the shop/company/… spells it. Only if they don’t use a Latin script name, we could use the official transliteration law’s rules. It might be a good idea to repeat the Cyrillic name in the name:bg=* tag for this reason.
If the name is in Latin script, it’s good to add the Cyrillic name in the name:bg=* tag for Bulgarian users that like to read it in Cyrillic. Spelling should be as in use by the shop/company/… or if there is no example, as you seem fit (I don’t think there are rules for it). It’s usually not necessary to repeat the Latin script name.
Some of the existing entries in the NSI are not in agreement with this. One is
the entry for Аптека Медея. It has name:en=Medeya included in it. I think we can’t determine to which language the word Медея/Medeya/Medea belongs to, so we should avoid assigning it to a language. It should be in int_name=Medea, in the spelling used by the company (in its web domain, Facebook user name, …). The same for Ремедиум…
Efbet uses the name “efbet” so that should be in name=efbet. The NSI entry also has name:en=efbet which I think is not correct (“efbet” is not an English word) and not necessary. Same for “Kaufland”, “UniCredit Bulbank”, …
“A1” is a special case because you can’t see if the name is in Cyrillic or Latin script. I think it’s good to add both “A1” (A in Latin script, in int_name) and “А1” (A in Cyrillic script, in name:bg) because search results might be different with your Cyrillic or Latin keyboard on.