On Fri, 10 Aug 2012 10:39:16 +0200, Roman Rokytskyy <roman@...>
wrote:

> Ok, we can switch to the new names, but switching to Charset classes
> should be checked carefully - the reason for the custom implementation
> is that gave 3-5% win compared to "original" Java. These 5% should not
> be thrown away just for the sake of using canonical implementations.

But was that tested against Charset (which was introduced in Java 1.4
together with NIO), or the mechanism used in Java 1.3 or earlier? As far as
I am aware Sun (and Oracle) have spent a lot of time optimizing Charset.

Mark