[9fans] off topic: full unicode support isn't easy

[9fans] off topic: full unicode support isn't easy

[Aharon Robbins]

FYI

Creative usernames and Spotify account hijacking
http://labs.spotify.com/2013/06/18/creative-usernames/


Arnold

Re: [9fans] off topic: full unicode support isn't easy

[erik quanstrom]

On Tue Jul 23 23:16:14 EDT 2013, arnold@skeeve.com wrote:
> FYI
> 
> Creative usernames and Spotify account hijacking
> http://labs.spotify.com/2013/06/18/creative-usernames/

i took a different lessons.  

0.  uncode support isn't binary.  one might reasonably support a subset.
(i wonder if they'd let me have “0021bf upwards harpoon with barb leftwards”
as a user name?  would it serve any business purpose?)

1.  it's important to write down what the requirements really are especially
when relying on external code to get it done.  better yet, have the code check the
requirements are met.

- erik

Re: [9fans] off topic: full unicode support isn't easy

[Kurt H Maier]

On Wed, Jul 24, 2013 at 05:59:32AM +0300, Aharon Robbins wrote:
> FYI
>
> Creative usernames and Spotify account hijacking
> http://labs.spotify.com/2013/06/18/creative-usernames/
>
>
> Arnold
>


Seems the thrust of this article is more like "writing secure
public-facing authentication mechanisms relying on unvetted clownsourced
code isn't easy".

khm

Re: [9fans] off topic: full unicode support isn't easy

[Steffen Daode Nurpmeso]

Kurt H Maier <khm-9@intma.in> wrote:
 |On Wed, Jul 24, 2013 at 05:59:32AM +0300, Aharon Robbins wrote:
 |> FYI
 |> 
 |> Creative usernames and Spotify account hijacking
 |> http://labs.spotify.com/2013/06/18/creative-usernames/
 |> 
 |> Arnold
 |
 |Seems the thrust of this article is more like "writing secure
 |public-facing authentication mechanisms relying on unvetted clownsourced
 |code isn't easy".

I thought more in the direction you-are-fucked-up-if-you-trust-
anyone-but-yourself-if-not-today-then-tomorrow.
(It's Python...)

 |khm

--steffen

Re: [9fans] off topic: full unicode support isn't easy

[Steffen Daode Nurpmeso]

erik quanstrom <quanstro@quanstro.net> wrote:
 |0.  uncode support isn't binary.  one might reasonably support a subset.

But it is not that you are totally against supporting
decomposition that includes Korean ("Hangul Syllable
Decomposition", must be performed algorithmically)?
It's only missing yet.

 |- erik

--steffen

Re: [9fans] off topic: full unicode support isn't easy

[erik quanstrom]

On Wed Jul 24 17:39:45 EDT 2013, sdaoden@gmail.com wrote:
> erik quanstrom <quanstro@quanstro.net> wrote:
>  |0.  uncode support isn't binary.  one might reasonably support a subset.
> 
> But it is not that you are totally against supporting
> decomposition that includes Korean ("Hangul Syllable
> Decomposition", must be performed algorithmically)?
> It's only missing yet.

i wasn't making any statements about particular languages,
or even endorsing supporting a subset.

but consider, file names in plan 9 support a subset of unicode.
they may not contain the codepoints u+00002f (/) or u+000000.
otherwise, a directory could contain the file entry "a/b" or "a/b\0c"
which could lead to confusion.  many file servers expand
the frogs list to include codepoints < or ≤ u+00020 (space)
to prevent accidents.

http://xkcd.com/327/

- erik