Re: man

[Phil Edwards <pedwards at disaster dot jaj dot com>]

> this's not a real reason, since the same thing have to do with the library
> itself:-)

Not really.  We provide to the end user a set of headers with declarations
and a compiled library containing the definitions.  The SGI implementation
is entirely a set of headers.  It's pretty easy to tell when somebody
is using (for example) the SGI version but complaining to the GCC lists.
It rarely happens.  Documentation is not so self-identifying.


> no. I expect two things. first something sgi's html pages about the
> standard library, second the or at least some minimal documentation about
> the extensions and not just about sgi's extension, but files like
> type_traits.h

If people want to read SGI's pages, then they can read SGI's pages.
They're well written, and -- more to the point -- SGI can keep SGI's
pages updated better than we can keep SGI's pages updated.

I agree that we should document GNU extensions, and at least provide a
pointer to SGI extensions which we make heavy use of in the library.


> and *concepts* which is not in the standard (yet) but very
> useful and without any documentation the only way to find this to read

SGI's concepts checks were added to the library three days ago.  I've
already written up some minimal documentation during lunch.  Give us
some time...


Phil