Since bug JDK-8061777 there is a performance regression for non-UTF-8 charsets.
This could be additionally painful when the creation of ZipPaths becomes looped by ZipFileSystem#iteratorOf().
At least, for single-byte charsets, the former performance could be regained simply by replacing ZipCoder.isUtf8 by something like ZipCoder.isSingleByte.
Further more improvements are possible, please follow: http://mail.openjdk.java.net/pipermail/core-libs-dev/2016-May/041559.html
Furthermore:
- consider subclassing ZipCoder from j.n.c.Charset
- consider benefit from bug JDK-6695386