Re: Rot13 and letters with accents

On 06/12/2007, John H. Jenkins <[email protected]> wrote:
>
> What you should use is rot32768 (applied on a per-plane basis).  Since
> the BMP and Plane 2 are both pretty full, most actual text would at
> least not contain large numbers of undefined characters (and Latin
> text would largely turn into Chinese, of course).  :-)

And some Chinese text would turn into text that includes unpaired high
and low surrogates as well as noncharacter codepoints. So, in my
opinion a Unicode version of a rot13 type obfuscator should apply
obfuscation on a per-block basis (and even then Arabic presentation
Forma-A would be problematic). Of course that would mean that you
would need to update the algorithm every time a new version of Unicode
was released, and so text obfuscated with one version may not be
correctly decoded with another version.

Andrew