From: Sean O'Dell <sean@...>
Date: 2004-05-29T05:29:24+09:00
Subject: Re: Calling global method

On Friday 28 May 2004 13:02, Austin Ziegler wrote:
> >> Does Ruby *need* globals, especially given that Kernel already
> >> exists?
> >
> > Kernel is just another way of saying global. How you implement
> > globals doesn't change the concept of globals itself. Whether
> > they're held in a nameless or named space, they're still globals.
>
> *shrug* I find that, with the exception of three methods I use
> almost as keywords (block_given?, puts, raise), I don't tend to use
> what's in Kernel.

You don't use eval, exec, exit, fork, load, p, proc, require, sleep, system, 
throw or catch?  I use all of those fairly often.

> > But it seems to me that Kernel ended up being the global namespace
> > after a couple paradigm changes, and there isn't much consistency
> > about them. Some globals are private members of Object and some
> > are in Kernel. You can access global variables anywhere using $,
> > but you can't access all global methods from anywhere. You can
> > access Kernel::method globals, if they're in Kernel, but if
> > they're a private member of Object, you have to perform some
> > strange tricks.
>
> There is consistency, though. If it's defined as a method in Kernel,
> then it's available "publicly" through Kernel. If it's defined by
> you, the user, then it's a private member method of the Object
> class. Since the execution context of all programs is through an
> instance of Object, those items that look like globals aren't.

Public through Kernel, private through Object.  That's not consistent to me.  
That sounds like a series of decisions that were made without much thought to 
previous decisions about where globals go.  Of course, it's tempting to say 
that it's designed this way on purpose, but stepping back and looking at the 
design as a whole, it looks kludgy.

> > It would be nice of all globals were wrapped up in a Global
> > module, so you could access them directly from anywhere. Variables
> > could still be sugared with $, and methods could be sugared with
> >
> > ::.
>
> *shrug* Again, I don't find the idea useful.

Well, I find it useful as I'm a bit stuck right now about how to call a global 
method without producing some ugly code.  It looks like ugly is going to win, 
though, since there is no clean built-in way to do it.

	Sean O'Dell