From: Herve Bronnimann (hbr_at_[hidden])
Date: 2002-09-10 21:02:14
On Tue, Sep 10, 2002 at 05:53:07PM -0400, Douglas Gregor wrote:
> After further consideration, I think this was the wrong decision. While I
> still believe that comparison policies are harmful, it seems that we do have
> a solution that retains the essential abilities of comparison policies
> without unnecessary danger.
In short, we all liked your proposal because it preserves the same
purpose as comparison policies, with the safeguards. One of us didn't
like the style (i.e. the C++ expression of it), but didn't have any
technical argument against it. And we can all live with it :)
By the way, thank you again (and the other reviewers) for the discussion
on comparisons and for going beyond the `it doesn't work for me', to
finding alternative solutions, and taking the time. We wouldn't have
gotten there alone!
> Therefore, I'd like to change my vote to
> "accept", conditional on one thing: I'd ask that the authors carefully
> consider the existing proposals for comparison policies and document
> (verbosely!) the reason for whatever decision they make.
We're definitely leaning toward the solution you proposed, and will
review it thoroughly, and implement it, and if we don't find unexpected
problem with it, it's more than likely it will be the comparison
mechanism of the library. If there had to be an unexpected problem with
it, we'd get back to you and the mailing list to try and fix it.
> I think this issue
> will probably come up again, and it would be very helpful to have a document
> weighing the various pros/cons of each approach to multiple comparison
> operator semantics.
Agreed 100%. I think it'll be interesting even to go over the discussion
once more and synthesize the arguments. And I had a feeling the
rationale needded to be beefed up anyway.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk