Re: pattern for implementation-private constants?

Quoting [e-mail], on Tue, May 08, 2007 at 01:21:56PM +0900:
> Somehow, just not documenting the constants leaves me unsatisfied, 
> though I admit that this is more of a theoretical than a practical matter.
> 
> Given that constants aren't actually required to be constant, anyway, 
> I'm tempted to use class variables for the private ones instead...  In 
> his response, murphy was kind enough to provide benchmarks showing the 
> performance hit I'd take if I did this.

I suspect that if constants could be private, it would introduce lookups
and checks that would give you similar performance hits. Fast and cheap,
or private, hard to get both at the same time?

Sam