the monopoly of classes

Cees de Groot cg at cdegroot.com
Fri May 23 13:04:02 UTC 2003


On Fri, 2003-05-23 at 12:23, jan ziak wrote:
> i disagree with you that Self is a purely classless OO language, because each 
> object has its parent (a field pointing to some parental object). the 
> parental reference seems to me to be very similar to "super" (which points to 
> the super-class of an object) in smalltalk.
> 
Disagreeing seems to be your bread and butter, not? The parental
reference can be nilled. No classes in sight. It's classless. So, Self
(and ECMAscript) uses prototypes. Why do you think they'd invented that?
For efficiency reasons.

> you are completely wrong in conjunction with the "inefficiency"  of genetic 
> code i think. the recurrence of dna sequences in different individuals has 
> its sense: it allows those individuals to communicate !!! it's not 
> inefficiency, it's natures magic behind that and the, according to your 
> reaction, "waste" has it's logic.
> 
Now, this is complete and utter nonsense and totally ignores what I
said. 

Please go off and count the number of copies of DNA you have (at least
that'll keep you quiet for the next aeon or two ;-)). That's what you
get with a totally classless system - a gross waste of resources.

There's nothing 'magical' about nature. It's an imperfect system,
created by evolution, and it shows. 

> the reason that you cannot communicate with an ape is that the genetic code 
> which decribes the ape is DIFFERENT from the genetic code of you body.
> 
Oh, and when you are done counting your cells, please count the number
of genes you have in common with an ape, and the number of genes you
don't have in common. I think you're in for a surprise.

> replication of code physically throughout your 
> body, of *exactly the same code* throughout the bodies of countless other 
> living things has its sense, it's fully rational and meaningful. (as einstein 
> said: "god does not play dice").
> 
> (note: don't you want to read 'A Mathematical Theory of Communication' by 
> Claude E. Shannon ? i think it would help you to take the same view as i 
> have).
>
Sorry, I don't want to take your view - if only because I don't follow
you.

Let me try again - I hope you can come up with more nonsensical gems
like quoting Einstein on quantum mechanics (which, incidentally,
probably says more about Einstein's lack of understanding - willful or
by some perceptual blindness - of Heisenberg's uncertainty principle
than about quantum mechanics) in a discussion on information stored on
biological carries. 

Because your cells form a classless system, you have *exactly* the same
information copied something like a billion times throughout your body.
And every other living thing has copies of often the same information,
copied as many times as they have cells. Now, I don't know how many
copies of DNA float around on planet Earth (it's less than a Googol, but
huge), but it strikes me as a gross efficiency. For starters, try to
debug the code and then play an update to all copies ;-).

Now, why is this? Two reasons: the need for local access of the
information (in each cell - and I'm not starting about the copies inside
your ribozomes in each of your cells), and the lack of reflective
capabilities in the system called 'evolution'. Evolution cannot step
back and rethink a certain design (if we're into quoting books, I hope
you've read Dawkins). 

Well, except for performance reasons (I have a replica of lots of code
on each of the machines I run - it's called 'the operating system' and
it wastes lots of diskspace and causes lots of management headaches, but
it is faster), I don't see why we should attempt to mimick a system that
is admirable in its results, but quite imperfect by design. 

Neither did the people who built Smalltalk, that's why they *could*
decide that information could be stored 'off-site' (outside the object)
and they could reflect on a better design. 

Anyway, except for bringing the whole mailing list off-topic, what do
you want to say? This is about Squeak development - do you have anything
concrete in the space of 'problems with class systems' or 'a proposal
for a classless Squeak'?





More information about the Squeak-dev mailing list