Christian Schmidt

Sorry, the link to #2469 was wrong. Should have been https://github.com/sparklemotion/nokogiri/issues/2469. Nokogiri is only used in tests, so there is no real security issue. But I agree that it is...

Thanks for the review. I agree about your proposed refactor, but I'd rather do that in a separate PR in order to keep things separate.

@flavorjones Thanks for your insights. Note that for `rel` attributes, we need to do a case-insensitive match for `me` in a space-separated tokens, so a pure XPath 1 solution would...

I think the use of custom handlers make the code easier to read rather than trying to achieve the same using the limited features of XPath 1. Upgrading to XPath...

> The change is welcome, although I'd maybe restrict the `source_language` param to known languages or just `'und'` and `nil`. The source language is checked against available languages in [`TranslateStatusService#permitted?`](https://github.com/mastodon/mastodon/blob/f6fb9b75dbc10cee8beee4bcb84dd08e821baf1b/app/services/translate_status_service.rb#L35),...

It is always displayed (after having pressed _Translate_ once), except when language tag is empty (in this case, auto-detection happens in any case).

I have tried to figure out what is going on. I don't fully grasp the complexity of the kt-mediaclip pipeline, but these are my findings so far: kt-paperclip and Mastodon...

Yes, that's it :-) The magic happens in the `Remotable` concern: https://github.com/mastodon/mastodon/blob/35683ac154ee7c717246dcac0d791b3f93c3824a/app/models/concerns/remotable.rb#L42

I have added a bunch of tests.

> And honestly, that last option would probably be good enough (at least for me). If false detections are a concern, then how about an option to to translate again...