Certain punctuation characters in BanConverter are defined to translate to XMLish <g> tags:
# punctuation ᭚ → '<g type="panti"/>'; # PANTI ᭛ → '<g type="pamada"/>'; # PAMADA ᭜ → \@; # WINDU ᭝ → '<g type="pamungkah"/>'; # CARIK PAMUNGKAH ᭞ → \,; # CARIK SIKI ᭟ → \,\,; # CARIK PAREREN ᭠ → '<g type="pameneng"/>'; # PAMENENG \u1B7D → '<g type="pantiLantang"/>'; \u1B7E → '<g type="pamadaLantang"/>'; EOF;
This doesn't actually work properly. In Parsoid, the tags are rendered as literal text, eg for:
https://ban.wikipedia.org/w/index.php?title=%E1%AC%93%E1%AC%A9%E1%AC%B8%E1%AC%A7%E1%AC%A2%E1%AC%BE%E1%AC%A6%E1%AD%84%E1%AC%A9%E1%AC%82%E1%AC%AE%E1%AC%B6&variant=ban-x-dharma&useparsoid=1&oldid=105938
we get:
Whereas the legacy parser's implementation of language converter actually opened a new <g> tag in the HTML:
I don't know what the proper transliteration of these punctuation marks are, but I'm pretty sure literal <g type="panti"/> is not the expected rendering.
This appears to have been present since the first revision of BanConverter in 9cb51879444cd5a72c60eec9b30a37c590e2330c for T263082: add <langconvert> parser tag.