> Why NONE encoding is evil ?

1) Because that means you can only have byte-order sorting on server.
That is okay for some alphabets but not for over.
Case insensitive search/sorting would not work either.

2) Because it is still not UTF-16 used by Java and someone somewhere would have to do the conversion.
So imaging you made some Java application, that needs to work with russian text and runs it
a) on Windows, using GUI
b) on Windows, using command line/scripting
c) on freshly created user account on Linux box
d) on Linux/BSD account installed decade ago and smoothly upgraded ever since.

Fat chance that you would have 4 different ways to convert same UTF-16 russian string to OS-local (and user-local) encoding of choice.
The same application (intranet site applet for example) would record different data for the same input, depending upon which client setup was used to invoke it.

> It is stupid to charge especially servers for encoding and
> decoding charsets.

I don't think so. Since most servers tun on HDD, the disk throuput would be bottleneck, not CPU-local transcoding. Especially if clients are little power devices or LAN connection is slow.