Re: a new language, designed for safety !

[Maciej Sobczak <see.my.homepage@gmail.com>]

> > What if we treat the standard library as part of the language?
> 
> It would be self-contradictory.

No. There are languages where even things like package importing, adding fields to record types or even virtual dispatch are library-level operations. They might or might not have a support at the level of what we consider "syntax", but even there the border is blurred. In languages that are somewhat inspired by Lisp even the for loop might or might not be a library feature.

> A library is defined in the language terms,

You are trying to give your own definitions as if they are universal truth. My definition is: language and its standard library are indivisible. What now?

I agree that it is a cool ideal to have a language that allows you to implement its own run-time or at least as much of it as possible. But I don't accept it as an objective definition.

> >> If Ada were a better language, array could be a
> >> library container.
> 
> > Assembly language fits that ideal, right?
> 
> No it does not. Assembly languages do not have containers.

You have asked for a library. Definitely assembly languages can have libraries with containers.

> Operator new was a mistake, excusable for early 80's. It is especially
> obvious since deallocator is not an operator.

Ada is exceptional in the sense that deallocator is not an operator. Other languages like have that right (that is, allocation/deallocations are both operators or neither one is).

> BTW, assignment operator was
> a mistake too.

If there is no assignment operator, you need something else to build it upon. A dictionary, perhaps? But now, you wanted that in the library, too.
It seems to be that the choice between syntax and library support is a matter of taste - but this makes it even further from being an objective definition.

-- 
Maciej Sobczak * http://www.msobczak.com * http://www.inspirel.com