Le 8/5/2005, "Yukihiro Matsumoto" <matz / ruby-lang.org> a ñÄrit: >Ilias, > >In message "Re: [RCR] Object#inside_metaclass?" > on Sun, 8 May 2005 04:09:28 +0900, Ilias Lazaridis <ilias / lazaridis.com> writes: > >|> If "singleton class" exists, it has an >|> instance. It would be an object of which the singleton class holds >|> attributes. >| >|But this object is _not_ an instance of the "singleton class". > >Not, in your diagram and definition, and not, in my explained language >model. But virtually everyone except us (me and you, Ilias) seem to >have the other model in mind. The object is an instance of the >singleton class in that model. It is also in the current >implementation of the interpreter. I'm thinking of changing the >definition to confirm model in others' mind, since the other model. >Under the new model, horizontal arrows in the object.c diagram mean >instance-of relations. People have a tendency to find complexity where there is none :) As far as a new model goes, I would assume everyone will be happy as long as one can metaprogram as easily as currently. >The obstacle is that if I choose the new model, I have to name an >official name to "singleton class", which I agree is not the best term >for it. > >|> |You've integrated existing concepts [which have existing terminology]. >|> >|> Then tell me the "existing" term for this concept, which I failed to >|> find in the past. >| >|I don't know it yet. > >Then please come back again when you find the name, preferably >suitable under the new model. Etymologically, 'idioclass' (as in an 'idiom') is probably the most suitable choice [1]. There was also a poll of some sort on Rubyforge, I think, if you want to see folks weighing in on their choice. > matz. E [1] http://www.etymology.com -- template<typename duck> void quack(duck& d) { d.quack(); }