strip initial exclamation and question marks in permalinks

[pandem]

These two symbols ¡ and ¿ are very common in Spanish titles, but they are not stripped out for the post slug. A similar bug has been reported and patched for 2.0.5 (​http://trac.wordpress.org/ticket/2735), so this one could probably be fixed for the same release.

[foolswisdom]

From 4636, %c2% is left in the permalink in place of ¿

[pishmishy]

Going by foolswisdom's comment this can be closed.

[lloydbudd]

Based on reading #2735, I think the ticket suggests that these should be stripped a together, and this issue isn't fixed.

[melado87]

The problem is that the ¡ and ¿ characters should be stripped out, just like the ! and ? already are. These characters are used in Spanish to start a question or an exclamation.

[pishmishy]

Should we change the behavior to white list allowed characters rather than playing whack-a-mole every time someone points out a character that's 'undesirable' in URLs?

[ryan]

Wouldn't we have to whitelist most of UTF-8? That's a big list. :-)

There's also the problem that changing what is stripped will break existing slugs made using the old code.

[pishmishy]

I hoped I'd get that response ;-) I don't think the whitelist needs to be that large. To work, whitelist would limit the slug to just (see RFC 3986 Appendix A)

ALPHA / DIGIT / "-" / "." / "_" / "~"

I'm not sure that'd be well received but if we could find a way to make it work.
I think we have three options.

• Do nothing. This seems inconsistent, we have done something for other characters.

• Whitelist the small set of characters that won't be percent encoded according to RFC 3986

• Whack-a-mole filtering of characters as and when people ask for it.

[westi]

Pushing to 2.9

Changing the list of allowable chars breaks old slugs.

There are a number of tickets for this issue for different characters.

We need someone to find all the relavent tickets and come up with a scheme for handling old slugs.

[mrmist]

Also related #5554 #4739 vaguely #6106 #1762 #4328 #6973

[Denis-de-Bernardy]

Merging into #9591.