Re: AVL vs Red-Black tree

[Paolo Carlini <pcarlini at suse dot de>]

Daniel K. O. wrote:

> Sure. But ABI breakage could be avoided by simply changing the body of
> the functions in tree.cc; unless of course the extra bit required by
> an AVL tree node could cause some problem.

I don't think things are so simple, unfortunately. Remember that when we 
say "ABI" in this list we mean something rather general: for example we 
want to be able to pass a map constructed by code compiled with the 
current library to code compiled with the new one. I don't think this is 
possible, right?

> Now that's some interesting point: what's considered a "large,
> significant contribution"?
> I wrote all the code (that is, 2 routines, plus some small changes to
> the others) around the libstdc++ implementation, so I would consider
> it a derived minor contribution.
> But that's no problem, I'm willing to do the paperwork if needed.

I would say it's certainly needed, and it takes time, a lot of time, 
unfortunately.

Paolo.