Re: Community Input for the Maintenance and Revision of the Ada Programming Language

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On 14/09/2017 08:57, G.B. wrote:
> On 13.09.17 21:34, Dmitry A. Kazakov wrote:
>> On 2017-09-13 19:39, G.B. wrote:
>>> On 13.09.17 09:30, Dmitry A. Kazakov wrote:
>>>> On 12/09/2017 23:04, G.B. wrote:
>>>>
>>>>>     X := new Message' ("Help me");
>>>>>     Y := new Messagio'("Aiuta me");
>>>>>
>>>>>     --  entering Little Italy:
>>>>>
>>>>>     X := Y;
>>>>
>>>> Shudder.
>>>>
>>>>> So, if you want to perpetuate the hazards of strings, ask the ARG
>>>>> for a ready made storage pool for strings.
>>>>
>>>> What is so specific about string storage pool?
>>>
>>> The fact that the pool will be there, doing its work without
>>> being noticed, to please those customers who don't like thinking
>>> about string handling and associated storage management.
>>
>> Which work?
> 
> OT: "The key to counted strings is that they need to be first class 
> objects.
>   The compiler must know how to handle them."

Both is wrong.

1. Bounded-length strings need not to be first class if the language 
type system capable of handling them. Ada's type system almost can, up 
to the moment when the user-defined object must expose array interface.

2. The compiler need not to know anything beyond user-provided 
implementation of the ADT.

> Thus, the work of handling strings. Strings in particular, because 
> apparently
> teachers think that they teach something valuable when they suggest
> using natural language strings or other semi-formal lumps of characters
> in a formal program's text.

Why these must be strings and why this specific kind of string?

Remember that bounded-length strings are in most cases premature 
optimization, because problem space bound either does not exist (e.g. 
non-static) or unknown. So the teacher is supposed to explain why X must 
be up to 100 characters long? Then the very idea of assigning strings is 
bad.

> Let's say that, seeing A. Perlis and A. Stepanov as professional
> mathematicians, there is an attitude---surely not theirs---that tries
> hard to make computation be classified information never to be
> shown to others. It is a mistake, though, to think that only the
> "result" of a computation is important, in that all of the
> computation affects "world". When and to what extent does this
> "extended result" affect {plan, system, supplier, customer}?
> Can you sell the effects?

You seem equate single act of computation and a process continuous 
computations, i.e. in a control loop.

Furthermore, there is a question of abstraction, At each level of 
abstraction there are exposed computational elements and hidden 
implementation-defined ones.

Anyway, if you consider computation so important for education, then you 
should not ask hiding memory management and issues of string assignment 
from the students.

> This is a challenging research topic for mathematicians specializing
> in programming:
> 
>   - to not exclude operations' effects from function specifications
>     (yes, specifications; it is partly done already when their
>      descriptions mention amortized cost)
> 
>   - to include more effects of operations when describing results.

OK

> Counted strings, then, serve the attitude that tries to indeed
> move education under the carpet (compiler), and for strings only.
> (Because, then, you don't _need_ to teach all those things that
> make counted strings work, and you move on to the _important_
> stuff...)

and in plain contradiction to the stated goals...

>> Furthermore your example
> 
> tries to allude to what might be a long term effect of counted
> string equivalents on professional programming if you use them
> in education, which you had explained further.

Yes, wrong paradigms have those effects. Not a reason to adhere to them.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de