IRI Encoding of slugs is broken

[hakre]

According to ryan, WordPress does "standard IRI-to-URI conversion for slugs." ​http://core.trac.wordpress.org/ticket/9480#comment:2

If that is true (what I assume), this feature is broken. Just tested the Post Title "Äpfel wachsen überirdisch" and it is converted to "apfel-wachsen-uberirdisch".

That is not the correct IRI encoding. Properly IRI encoded, the string must be converted to: "%c3%84pfel%20wachsen%20%c3%bcberirdisch".

[ryan]

We decompose accented Latin characters to their unaccented forms. See remove_accents().

[hakre]

Well, that's irritating. So infact the the slug is partly IRI encoded? Is next to remove_accents() other stuff made? Why is that done anyway, the IRI stuff looks very good.

[hakre]

Well obviously spaces are replaced with minus-signs as well.

[ryan]

It's done because people like it. Regardless, changing how slugs are encoded breaks resolution of old slugs. The current method won't change until we have a means of resolving slugs made with different versions of the sanitizer.

[hakre]

sanitizer = sanitize_title();

[hakre]

Related: #12361.

[texttheater]

One can live with this as long as it's possible to fix it via the plugin API. Which it isn't anymore in 3.1, see #16905.

[texttheater]

Sorry, I just wanted to add myself to Cc.