It would be better if the download of the langtags.json file could be separated out from the initialization code and allow applications to decide when to make the call.

@jwickberg Don't we have that already? There's Sldr.InitializeLanguageTags() which does the download of langtags.json. The application can decide when that gets called. If it has been called before (i.e. if _languageTags is not null) we won't do the download again. Or do you mean that we should always use the local langtags.json file, and do the download in a separate method that the application can call when it pleases to do so?

@ermshiperete I created this when I found that Paratext was starting slowly on some machines because of the download of the langtags.json file - this was made worse because the IF-MODIFIED-SINCE wasn't working which meant the file was always being downloaded. That was corrected on the server, but WSTech suggested we do the check a different way.
If we are sure that the timestamp check is working - the flag is less important.
Probably need to check with WSTech again to see the best way to make sure that the timestamp works.

I think what needs to happen:

1. extract downloading code to a new public method
2. change signature of Sldr.InitializeLanguageTags() to Sldr.InitializeLanguageTags(bool downloadLangTags = true) and call the new download method only if downloadLangTags is true. This deserves a +semver:major in the commit message (and an entry in CHANGELOG.md).
3. change download method to use ETag (IF-NONE-MATCH header) instead of IF-MODIFIED-SINCE header. This requires a way to store the ETag of the downloaded file somewhere, maybe in a separate file (langtags.json.etag). Downloading should re-populate _languageTags.