Name: dbT83986 Date: 02/10/99


The behavior of String.getBytes changed
between 1.1.6 and 1.1.7. In 1.1.6 and below,
it simply returned the low byte of each
character. In 1.1.7, many of the characters
(128..159?) are converted to 63 (a question
mark).

I understand that getBytes is supposed to
convert characters to bytes based on the
default character encoding. But I think
the larger issue is whether methods like this
should change behavior within the 1.1.x
line of Java api's. I don't think they should.
If using just the low byte was a bug, you
should have fixed it in 1.1.1, or documented
the 'bug' and left it unchanged. Otherwise,
I think you should hold off making changes
until 1.2, if ever. I don't think the behavior
of a method should ever change. If you want
different behavior, make a new method
(for example, getBytesEnc()?). Otherwise,
you will just keep breaking programs.

More and more, I find myself writing my own
versions of Java methods to improve performance
and insulate my programs from changes and
bug in Java's methods. Otherwise, I can't
write programs that can be used by clients
with 1.1.6 and 1.1.7.

Please set a policy regarding this issue
within JavaSoft.
(Review ID: 47847)
======================================================================