[self-interest] Our two weapons are fear and surprise...and..

[Niko Schwarz]

Can I digress very briefly on this matter?

The natural numbers are actually constructed very simply and  
beautifully. Or can be constructed, I should say. You start by just  
sets, and define the empty set to be zero. Then you allow a set to  
contain other sets. So we define 1 to be the set that contains 0. Now  
we we define for each number, that the next number is the set  
containing all its predecessors.

That is a beautiful definition, and the building blocks are simple: We  
need only an empty set and sets that contain other sets. And we get  
meaningful, elegant definitions for things like "smaller  
than" (through set inclusion, in this example).

What I want to say is: Simple building blocks don't necessarily lead  
to a simple system. Knowing this, in mathematical writing, I am  
sometimes tempted to write n when I mean the set {1,…n-1}. Because I  
know it's defined to be this way. But then I realize that that would  
be queer, that the majority of the mathematicians has a very different  
mental model of natural numbers. Natural numbers are these things that  
define the quantities of finite sets. You ADD them, sometimes  
multiply, and they fit quite nicely into the Range of real numbers,  
rather isolated.

Now, you could argue that it is because that's how we get used to  
natural numbers. That's how we use them every day, if it were just  
taught differently, they would enjoy set tricks more. And I would  
reply that New Math failed: http://www.youtube.com/watch?v=tx5KDyvlG3Q.

Because, you see, building systems bottom-up makes them conceptually  
beautiful, powerful, sophisticated, and without rough edges. But they  
also ensure that things that are fundamental to the user are really  
quite involved concepts in your system. I mean, it does irritate  
mathematicians if you use "element of" and mean "smaller than" on  
natural numbers. Eventhough it is, in a sense, a fundamentally  
equivalent thing.

The point I am getting to is, a system like Self might be beautifully  
simple, because the design principles can be taught in less than a  
minute. But the newbie does not NEED the design principles, he needs a  
for loop. Now, given Self, he has to learn two things: the for loop  
AND the design principles.

I quite understand where the OP is coming from. I've taught a few  
friends smalltalk, and I was ASTONISHED to see how long it took to get  
people to grasp it. Not at all less time than Java, I would estimate.  
Despite the fact that Smalltalk is, conceptually, a lot easier than  
Java.

Systems like Self are very easy for people like us. We know the for  
loop concept, the closure concept, and the accessor concept. Self has  
an elegant way to implement these, so for US Self is heaven. A good  
computer scientist can probably learn Self in a day, because he knows  
all the concepts already. For US, it really boils down to only  
understanding the few fundamental principles of Self, and the rest  
follows easily.

Maybe to wrap up, just as a case in point: in the parameterized  
complexity field of mathematics, which develops algorithms in large  
quantities, the usual pseudo-code for algorithmsis is imperative.

So, my point would be that there is a certain simplicity in Pascal,  
that you cannot find in Self. Pascal offers a syntax piece for each  
core concept that you really need to know. And not more. Not even a  
break statement.

cheers,

Niko



On 05.08.2009, at 20:46, Jeremiah S. Junken wrote:

> It may have something to do with familiarity. Self and IO are both  
> extremely simple, and for people familiar with either one, learning  
> the other is fairly simple -- but for people who have not started  
> out with something like Smalltalk, or a functional language, it's  
> definitely a bit of an alien landscape with a learning curve
>
> There's conceptual simplicity and practical simplicity -- the latter  
> has a lot to do with familiarity.
>
> My 2 cents in debased currency.
>
> On Wed, 2009-08-05 at 10:59 -0700, Randy Smith wrote:
>
>
>> I'm pleased you say "I agree that the language it simple -
>> semantically" as that was the intended claim. Maybe we should put
>> that in the title, but it makes it little clunky sounding.
>>
> [snip]
> [snip -- pwd vs. self ]
>>
>> I suppose the larger question might be: Every language is like a
>> carpet with a complexity bump that can be moved around from place to
>> place but never flattened out. Are some language carpets  
>> intrinsically
>> less bumpy?
>>
>> --Randy
>
>
>
>