Re: objective-c: how slow ?

[Marko Mikulicic]

Richard Frith-Macdonald wrote:
> So, with delegation and object A which does not handle a particular
> message, passes the message to a delegate object B which executes
> the appropriate method, whereas with behaviors, A gets the method from
> B and executes it itsself.

Sorry, its only a fact of terminology. I mistaken calling it "true"
as opposed to "not true". I'm grown up with this termin, and many people use it,
but it could be wrong or at least have different meanings.

> The behavior mechanism is not directly supported by the language, and
> is therefore dangerous in ObjC - since you need to be careful not to
> take on any behavior which would try to work with instance variables
> that don't exist in your object and cause a crash.  In other languages
> where ivars are only accessed by methods, you just need to worry about
> runtime exceptions, not actual crashes when you make a mistake.

I have a vision about OO language which certainly you won't share (You would be 
probably using smalltalk not objc), that instance variable must not be accessed 
directly, so that subclasses can overrive ivar access with some method. (for 
flexibility). The VM/runtime should optimize out actual unnecessary calls, but 
there must be the flexibility.
 I don't want to dispute OO desing here. I think I would not be using objc if 
there was not OpenStep. I think that NeXT guys created a very confortable 
framework (althrough when I first saw these loong method names I felt pain) 
where you can be very productive from the beginning. And objc doesn't need any 
new feature for OpenStep.

Marko