On Tue, Feb 05, 2008 at 07:13:40PM +0900, Jacques Garrigue wrote:
> > that's really a bad idea; As a user of xmlm, I hope you're going to
> > re-consider. the polymorphic variant namespace is so easily polluted by
> > random "value" that library should never use them or at least doesn't
> > advertise them as public interface.
> 
> I have no particular opinion on this particular case (if you want to
> allow chaning the library, you can also functorize your code, which
> would work with normal variants too), but could you explain how
> polymorphic variant namespace can get polluted?

I consider them as pollution since they don't have a "namespace", but that
might not be the right word here.

What i want to express is the fact that when I use "`Node", i can't tell
from where it's coming from;
is that an Xml node, a binary tree node, a Red/Black tree node, etc ? dunno.

when I have a "Xml.Node", i automaticly knows that Node is coming from
the Xml namespace, and it's an xml node.

> The point of polymorphic variants is precisely that pollution does
> not exist (i.e. only constructors that appear in the same type
> matter). This is what makes them so nice in libraries.

I beg to differ. I'm happy to know from where my symbol comes from,
since that's automatic code documentation for me. using polymorphic
variant just put everything in the same flat namespace.

-- 
Vincent Hanquez