[boost] Re: [shared_ptr] PPC implementations updated

From: Miro Jurisic (macdev_at_[hidden])
Date: 2005-04-08 21:46:44

In article <4256FADB.186931E7_at_[hidden]>, Alexander Terekhov <terekhov_at_[hidden]> 
wrote:

> Miro Jurisic wrote: [...]
> > I think that given the choice of better performance and better safety we 
> > should err on the side of safety until the faster version is provable safe, 
> > not the other way around (as you seem to be suggesting).
> 
> Faster version doesn't become unsafe merely because a preliminary notice 
> "strongly recommends" some silliness not giving any reason for "strong 
> recommendation" whatsoever. 

Unless you can provide some credible data showing that the strong recommendation 
by chip manufacturer is safe to ignore, it would be prudent to follow the 
recommendation.

What is it that you know that we don't? So far you have asserted a variety of 
things, all of which boil down to you being an authority on this subject -- 
including saying above that we should disregard an explicit strong 
recommendation from the chip manufacturer -- but unless you can provide some 
reason why your assertions are valid (preferably in the form of data, but 
credentials and references will do in a pinch), your argument doesn't really 
hold much credibility compared to official documentation by IBM.

I'd be happy for you to be right, for this issue to be moot, and for us to be 
able to use the faster code. Please give us a solid reason to believe that the 
IBM documentation under discussion is wrong (other than "I say so").

meeroh