Michael Link wrote:
peter royal wrote:
Currently, ByteBuffer allocation is done via static methods. The way
to do that would be to remove the static methods, and pass around the
ByteBufferAllocator anywhere that a ByteBuffer is needed.. which will
trickle around to many places in the code, alas.
If we were to do that, pre-1.0 is certainly the time.. anyone else
have thoughts on this?
I think this is quite a low-level configuration like the ones (-X...)
of the JVM so it shouldn't clutter the high-level API und thus make it
harder to use. I don't understand why the configuration of this flag
is that much different in an IoC-Container. I don't know Spring well
enough, but in HiveMind this can easily be done in a clear way via
configuration points. So I would vote for keeping the static methods.
I was thinking of the case when one application runs multiple TCP
servers, possibly supporting different protocols, and some protocols
benefit from direct buffers and others don't. Anyway, it's no big deal,
just an orientation I like to maintain where possible in my own stuff.
-0xe1a
