The discussion centers on disallowing RTL (right-to-left) override characters in topic titles due to rendering and security issues. locallunatic highlights that these characters break UI elements like the edit pencil and title indicators, and notes that the issue arises from mixing RTL/LTR text in titles [1]. codinghorror compares this to general Unicode griefing and suggests blocking or suspending users who abuse it. locallunatic adds that the edit pencil behavior is also affected, and recommends blocking such characters rather than allowing them. haiku points out that these characters also impact how titles are stored in the database. sam proposes allowing admins to use regex filters and a site setting to handle exceptions, noting that maintaining a comprehensive Unicode whitelist is impractical. locallunatic counters that the Unicode itself isn’t the problem — it’s the mixing with LTR text — and suggests isolating the title text from UI elements like the edit pencil and new topic bubble. sam agrees that adding an extra span for rendering separation is a viable, low-effort fix. The consensus leans toward preventing RTL override characters in titles via filtering or separation to preserve UI integrity and avoid abuse.
The new indicator for topics has problems with things that are allowed in the topic title.
Note: this is from http://what.thedailywtf.com/category/meta/one-post and looks like it is due to unicode support in the topic titles (at least that is where the NEW is showing).
Edit: specifically the LTR and RTL overrides.
No different than any other kind of odd Unicode content griefing and should be handled the same way.
I’d just block or suspend any user who persisted in doing this for now.
Possible that we could disallow certain unicode characters from titles, but this is playing whack-a-mole.
It also plays with the edit pencil on the title when you do it. So yeah it is an issue with allowing unicode characters, but don’t know if others would also break it. I’d suggest using a whitelist rather than a blacklist.
It also affects how the titles are stored in the database
I am ok with allowing admin to pass it through a regex and adding a site setting.
Building a whitelist of every unicode that is valid and known to man is not easy, also zalgotext may be “valid” in your local its just that its overdoing it.
I don’t think this battle can easily be beaten by computers. Its why you have a flag button.
Really there is no problem with allowing the unicode in the title, you just need to separate the title from where the NEW alert bubble and edit pencil go. Surround the title in span or whatever and put them next to that.
Yeah an extra span there is important for styling anyway.
