[TYPO3-core] Question regarding character conversion support

[Christian Kuhn]

Hey.

I'm currently thinking about a "auto configure" system in the install 
tool that aims to set sane defaults / suggestions based on system 
environment.

In this process I again stumbled upon our charset conversion (formerly 
t3lib_cs) class. Currently, it implements 4 solutions: mbstring, iconv, 
recode and 'own code', defaulting to 'own code'. The default solution is 
extremely slow, some different default should be set.

For the last couple of years I didn't see any PHP without mbstring in 
the wild. Even debian lenny (php 5.2.6) delivered with it enabled in the 
base package.

So, my question is if we could just rely on mbstring and drop the rest? 
Is it ok if we expect mbstring to be available? Are there *common* 
setups without enabled mbstring? Question is towards usual linux 
distros, windows and mac systems, but not to handcrafted systems like 
gentoo.

My goal is to either fully rely on mbstring (it the quickest one), or to 
enforce mbstring or iconv (recode is broken anyway). At least I would 
like to get rid of the 'own slow code' solution. Is that ok?


Regards
Christian