handling of unknown languages in \DocumentMetadata

[u-fischer]

If a language that is not known to babel is set in \DocumentMetadata it errors. It doesn't matter if the language tag actually exists (in https://www.iana.org/assignments/language-subtag-registry/language-subtag-registry) or is completely wrong:

\DocumentMetadata{lang=ht} % or ff or blub
\documentclass{article}
\usepackage{babel}
\begin{document}
abc
\end{document}

gives

! LaTeX Error: Missing \begin{document}.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help.
 ...rmessage  LaTeX Error: Missing \protect \begin {document}.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help\@err@                                                                   
                                                                                                                                                                                                                                      
l.4115       \bbl@read@ini{\bbl@bcp}\m@ne

[jbezos]

​babel doesn’t validate the tag in \DocumentMetadata. See the explanation here. Although babel can do an additional check, imo this should be detected even if babel is not loaded, or otherwise the PDF will contain invalid metadata.

The error is cryptic, indeed. For the moment a valid tag is assumed. It will be sorted out soon.

[u-fischer]

But ht or ff are valid tags, you find them in the registry linked above, only blub is invalid.

[jbezos]

I think loading babel for an unsupported language doesn’t make much sense. We can define fallbacks, but there are more than 8000 entries with type ‘language’! Even the Unicode CLDR falls short of 1000 languages.

[u-fischer]

Well you don't know why babel is used and if it makes sense, it could be e.g. for a second language.

babel doesn't need to validate the value of lang, \DocumentMetadata doesn't do that either. But imho if it doesn't know the language it shouldn't error but ignore the value (perhaps with a warning).

[jbezos]

Point taken.

[jbezos]

Since the ini file contains information relevant to babel, some additional actions from the user would be necessary. For example, the OpenType language system (at least the script) may be essential. This means a simple fallback to en (in full en-Latn-US, according to the likely tag) from, say, shn (Shan, in full shn-Mymr-MM) doesn’t seem sensible, and sooner or later an error may occur.

The next step should be to decide what to do when the simple lookup removing, if necessary, the script or the region is not enough. But for the moment, and considering the number of documents in unsupported languages is low, a more explanatory error will be raised with a method to overcome it in a forward compatible way.