Towards the end of this article there is a reference to the 'metaclasss'. It should probably be noted that the prefered terminology in the Ruby community now is 'singleton class' and not 'metaclass'.<p>For quite a while there was little agreement on this name because there was no standard method that could be called to return the singleton class and therefore no agreed upon name. The only access was via syntax:<p><pre><code> (class <<object; self; end)
</code></pre>
The results of this expression were called metaclass, eigenclass, singleton class, and probably a few more names.<p>There is now a standard method:<p><pre><code> object.singleton_class
</code></pre>
and so, for the most part, the naming confusion is fading away.
This is one of the areas that prototype-based languages simplifies although access issues might trouble class lovers. I do miss NewtonScript for some things.<p>Example: I create an object called Polygon with the intent to use it as a prototype. I assign the variable sides with a value of 10. So, anything with a parent pointer of Polygon gets 10. I then create Triangle from Polygon and assign sides = 3. It doesn't affect Polygon (which has its own variable), but anything that has a parent pointer of Triangles get the 3.
I'd never read the linked Alan Kay paper about the development of Smalltalk [1]. It's fascinating!<p>I've been trying for awhile to find readings or videos (or anything, really) about those early days at ARPA and then PARC. Especially the detailed discussion of the smalltalk development. The names and stories are also pretty neat to read, and have already inspired a few ideas personally.<p>[1] <a href="http://www.smalltalk.org/smalltalk/TheEarlyHistoryOfSmalltalk_Abstract.html" rel="nofollow">http://www.smalltalk.org/smalltalk/TheEarlyHistoryOfSmalltal...</a>
This was also discussed five years ago in this great post by Martin Fowler: <a href="http://martinfowler.com/bliki/ClassInstanceVariable.html" rel="nofollow">http://martinfowler.com/bliki/ClassInstanceVariable.html</a>
In the Perl/Moose world, class variables (attributes) are regarded as <i>a broken concept</i>: <a href="http://www.perlmonks.org/?node_id=690482" rel="nofollow">http://www.perlmonks.org/?node_id=690482</a><p>ref: <i>stvn</i> == "Stevan Little" who is the creator of Moose (<a href="https://metacpan.org/module/Moose" rel="nofollow">https://metacpan.org/module/Moose</a>) & and it's MOP underpinnings (<a href="https://metacpan.org/module/Class::MOP" rel="nofollow">https://metacpan.org/module/Class::MOP</a>)
In Kay's Smalltalk, humans never saw code like that in the article: they defined classes, variables and methods through a graphical interface. The verbose syntax was for computers to read, when you saved code on one machine and loaded it on another. Humans only had to write this stuff after Gnu released a text-only interpreter; the subset of Smalltalk syntax that was originally intended for human use is pretty clear.