On Tue, Aug 28, 2007 at 02:33:23AM +0000, Angus Leeming wrote: > Andre Poenitz <[EMAIL PROTECTED]> writes: > > So anybody who remembers the reasons to use the anon namespace (for > > _functions_)? [Rhetorical question...] > > I seem to remember that Lars told us that it was "a good thing" (TM). > However, > before anyone indulges in a "let's bash Lars bit of nostalgia", here's what > Stroustrup has to say in "The C++ programming language", 3rd ed: > > Section 9.2. Linkage > > In C and older C++ programs, the keyword static is (confusingly) used > to mean "use internal linkage" (ChB.2.3). Don't use static except inside > functions (Ch 7.1.2) and classes (Ch 10.2.4).
Well, this does not really give a technical reason why namespaces should prefered. And I don't buy that "(confusingly)", even from Stroustrup. After all, static on file scope _means_ internal linkage, also in C++. Or are you aware of any other means to achieve the same? > So, whilst this certainly doesn't invalidate your experiment, I think > that we should also take a rain check and ask just how much does > following Stroustrup's advice really hurt in practice. > > If it costs 3 secs on a full compile, then any conclusions are > essentially worthless ;-) Why? To measure overall performance difference I'd have to try it out. After that, the work is already done and in case there were _any_ improvement (which I do not doubt right now) I do not see that we should stick to the namespace version just because BS said so. Andre'
