[Jaakko Jarvi]

for_each is the only one that takes a 'free-form' function object.

Yes, as I said, it is "special".

If side effects are allowed, I don't understand why for_each is classified as a non-modifying sequence. It sure can modify the elements of the sequence. What is (philosophically) different in the way that, say, std::replace, modifies the elements of the sequence?

Because for_each() does not do the modifying, the /functor/ does the modifying. Contrast replace(), which does the modifying itself.

Do you know what Bjarne is referring to with the 'recent standards resolution'?

I have E-mailed him asking for a clarification.

The sgi STL documentation (predecessor of what is currently in the standard)

... is wrong wrt the final Standard.

This must be the initial reason why for_each is in the category of non-modifying sequences.

It may have been that this was the intent pre-Standardization.

A DR would be in place: either explicitly mention that side-effects (of the sort that modify the values in the sequence) are not allowed, or move for_each to mutating algs. Volunteer to raise that and provide wording?

I don't believe that a DR is necessary given the phrasing in TC++PL, which is reasonably authoritative (though lacking Standard force). Stephan T. Lavavej