#7 When parsing non UTF-8 encoded XML the mark/reset fails

[Sven Java]

When parsing non UTF-8 encoded XML documents the
mark/reset fails (in our environment). The mark is placed
to reserve 71 bytes, the first reset() occurs after
reading only the initial 4 bytes, when doing a reset()
again after detecting a different encoding than UTF-8
(or rather, than the assumed encoding) there have been
8192 bytes read due to buffering. Our current
workaround is placing a mark for a bigger length.