Which way is better ?

That is fine. I am merely pointing out that you shouldn't blindly put using namespace statements for every namespace you encounter.
Especially phantomotap seemed to think that explicitly specifying namespaces was like the old C days when you had to spell it all out. Or at least it seemed that way.
I didn't agree, because I use namespaces for categorizing. And the using isn't evil; there are places where it helps, like in your example, and others where it does not.
Myself, if it is not too long, I do prefer to type it out explicitly.

I'm just saying one [...] not back in the C era.

O_o

You are confusing "may" with "must".

If one "may" fully qualify names, for whatever reason, it isn't bad at all.

If one "must" fully qualify names, the namespaces are being used solely as a prefix. All the advantages of namespaces are lost. There is no difference between one prefix notation and the other. The source may only use ':' instead of '_'.

Which is more readable?

Code:

std__copy(std__istream_iterator<std__string>(std__cin), std__istream_iterator<std__string>(), std__ostream_iterator<std__string>(std__cout, " "));

Code:

// using directives
copy(istream_iterator<string>(cin), istream_iterator<string>(), ostream_iterator<string>(cout, " "));

Soma

Originally Posted by whiteflags

Namespaces are there to resolve things when the same identifier has been used multiple times, and the above communicates to me that namespaces somehow lose their value when they fulfill their purpose.

I think you have made a strawman argument since phantomotap compared namespaces to name prefixes, and it is in that context that he stated that "all the advantages of namespaces are lost" (emphasis mine).

*shrug* I deleted my comment because it occurred to me that Soma was probably saying what I was saying, in so many words. That happens. I'd ignore him so that I don't bug him but he's actually smart, says great advice, and that's mean.

Originally Posted by phantomotap

O_o

You are confusing "may" with "must".

If one "may" fully qualify names, for whatever reason, it isn't bad at all.

If one "must" fully qualify names, the namespaces are being used solely as a prefix. All the advantages of namespaces are lost. There is no difference between one prefix notation and the other. The source may only use ':' instead of '_'.

Which is more readable?

Code:

std__copy(std__istream_iterator<std__string>(std__cin), std__istream_iterator<std__string>(), std__ostream_iterator<std__string>(std__cout, " "));

Code:

// using directives
copy(istream_iterator<string>(cin), istream_iterator<string>(), ostream_iterator<string>(cout, " "));

Soma

Ah, yes. You are right. Still, I prefer

Code:

std::copy(std::istream_iterator<std::string>(std::cin), std::istream_iterator<std::string>(), std::ostream_iterator<std::string>(std::cout, " "));

Not only do they avoid collisions (which is pretty irrelevant in this example, but still), but I can see where they come from, and I do like to prefix this explicitly, unless too long.
Now, those are good arguments to ponder. When to use using namespace directories and when not to?

I'd ignore him so that I don't bug

^_^

You shouldn't bother about that; everyone bugs me anyway.

When to use using namespace directories and when not to?

Beyond rules that prevent pollution or allow overload resolution, this seems to fall under the realm of style.

Soma