Talk:Unicode variable names: Difference between revisions
→Why!: fix broken link
(→Why!) |
Thundergnat (talk | contribs) (→Why!: fix broken link) |
||
| (3 intermediate revisions by 2 users not shown) | |||
Line 1:
==Why!==
I just saw [
:Yeah, that is not good at all. It means that the source code needs conversion to be usable on ascii based platforms. Maybe we will need a unidecoder for such sources. That is probably a future task. [[User:Markhobley|Markhobley]] 20:57, 6 July 2011 (UTC)
::The change was made to Perl 6 code, and it is my understanding that that's correct code for that language. Usage of Unicode glyphs in Perl 6 code is pretty normal. --[[User:Short Circuit|Michael Mol]] 21:58, 6 July 2011 (UTC)
Line 59:
:They could do with grouping together similar looking symbols, and making them interchangeable, so that substitution can occur. I was just looking at the Unicode table: (five identical looking pipe signs), પ and ૫, ඤ and ඥ, ረ and ሪ, ⬦ and ⬨. That was interesting. These last two appeared filled in one browser window, but hollowed out, when I pasted them here. Some symbols have no visual difference whatsoever. It is going to be a right nightmare trying to spot mismatched characters. [[User:Markhobley|Markhobley]] 01:12, 10 July 2011 (UTC)
:: They're not ''semantically'' interchangeable. They might not have the same glyph in all fonts. They might participate in ligatures differently (Latin-based writing systems are largely simple that way, but other writing systems are very much not). And anyway, it tends to not be such a huge problem in practice; individual (human) languages don't have to deal with the problem in the first place, it's only when you try to support all known writing systems that you run into problems. (If you want an area where there ''are'' problems due to the sorts of issues you mention, try unicode domain names; that's a totally different problem from source code though.) –[[User:Dkf|Donal Fellows]] 23:45, 10 July 2011 (UTC)
==Security considerations==
I want to add an overview which approach was taken:
* ASCII only
* None (such as php, D, nim, crystal, ...)
* Certain ranges (e.g. ALtId in C11) (C, C++) (e.g. prone to rtl attacks)
* only TR31 (Letters, Numbers, ...)
* disallow excluded and limited use scripts
* which security profile (2-6) esp. how to check mixed scripts,
* require normalization, and which. (such as the odd choice of NFKC with python 3)
I'm working on such an overview and tools for all main stream languages (and filesystems and everything with names). Because 99% of them are insecure and do not follow the unicode security guidelines for identifiers, leading to unidentifiable identifiers. https://github.com/rurban/libu8ident --[[User:ReiniUrban|ReiniUrban]] ([[User talk:ReiniUrban|talk]]) 09:34, 30 December 2021 (UTC)
| |||