Re: Invade wikipedia!

["Hibou57 (Yannick Duchêne)" <yannick_duchene@yahoo.fr>]

On 25 fév, 00:01, Georg Bauhaus <rm.tsoh.plus-
bug.bauh...@maps.futureapps.de> wrote:
> The word "precision" can be misused.  The precision of an algorithm
> is measured against what? More than some typical level?
> Less than it?  What's typical? And why? --  That is,
> should Wikipedia, in an algorithms book, be so precise
> that it illustrates certain CPU-specific integer overflow
> detection machanisms in the examples?[*]  Or should it delegate
> discussions of this level of precision to some book about
> implementation strategies?
As too many references are possible, I suppose this would be
delegated, using a formalism which would help to make the link between
these two references.

Algorithmic is oftenly learned using such abstractions.

> These constraints add real things, but they do not help with
> achieving the clearly specified goal (of explaining the
> algorithm): When I want to explain just how to sort a thousand
> arbitrarily large whole numbers there is no need to refer to
> the limits of a machine model.
I agree, this is an algo.

> Worse, I start waving hands:
> "You know, just assume these magic types I gave you can
> deal with large whole numbers fine. I'm omitting the
> details of how this is possible etc etc."  What you loose,
> then, is what you wanted to add: precise explanation,
> of nothing but the algorithm!
>
This is another step of the process, which deals with different
matters.

Previously I've said that an efficient or good enough implementation
is oftenly far from what looks a theoric algorithm... while this does
not mean the algorithm does not drive the imagination on the road to
the effective solution.

> > If this kind of pseudo code was suffiscient to really and precisely
> > express algorithms, it would be a real language.
>
> Pseudo code of the kind in question is a real language.
> By analogy, no one writes Turing machine programs in
> commercial applications[***]. Still, his fictional(!)
> machine is entirely programmed using "pseudo code"[****].
> A Turing machine is defined precisely and
> its languages are real programming languages.
When I read you, I'm thinking about something like LISP. Is that Ok if
got this image in mind ?

> The machine *may* be a sequential digital RAM of the kind we
> think we know. If this is important, it will be instructive to
> *add* more aspects to the algorithm.  Can we afford large numbers?
> What *is* a large number, by the way?  The a real programming
> language will become important insofar as it reflects
> the constraints imposed by the digital RAM and the PL definition.

This is a next step, applied to an algorithm, which may have numerous
live cycles (the algo).

> Properties of what? Try it, I discovered many details that
> need to be masterd before the expressions become clear.
> They are just not stated in the declarations, but in preparatory
> texts and preparatory exercises.
Properties like "can I compute this in a human life time ?", or else
"how does the computation time will increase ?"... these are fuzzy
view, I agree, I know this is many times a bit far from the reality,
while what it says is most of time true in the big shape.

> Algorithms being better discussed using declarative meanings is
> a claim frequently made (and hardly ever shown to be correct)
> by mathematicians.
> Statistically, there is a remarkable correlation between
> those who make this claim and their profession (mathematics).

I'm my self really out of this staticical assertion.

> When we try to understand a declaration or equational
> definition, our brain does operate(!)!  It does not just denote,
> if it does denote at all.
Yes, it operates, in its most convenient way of the moment, which is
far less restrictive that what would be any real machine. We operate,
but in a far differente way a machine would do. We get freedom we
would not have if we were operating like a machine.

> In order to really understand what a "functional equation"
> means, you have to perform(!) some backtracking:
>
>    f(0) = 1
>    f(t + 1) = "*"(t, f(t - 1))   for all t in N
>
> To understand what is going on here you need to follow the
> recursion. This process(!) is never emphasized in a
> functional showcase.[******] Why not?

You are right again..... but we do not follow the process like a
machine would do. We may sometime go using a shortcut (which may be
wrong by the way, thus making a wrong intial assertion) a machine will
never use (not so far, perhaps in the futur this gonna not be true
anymore).

> Proceding on the declarative route, you arrive at assembling(!)
> functions in tricky ways in order to achieve an efficient
> evaluation strategy. Isn't this totally obscuring the explanatory
> goal?


I apologize to not have replied to all exact points (I'm not as fluent
as you are at english), but any way, I would like to temporaly end
with this (as least, an attempt) : What've learned so fat, is that an
algorithm as least something similar with a piece of code (supposed we
all agree). Peoples may write this kind of piece of code with
different goals in minds (as peoples are all different and has
differents though). Then, one see this, and try to map it on its own
expectations.... finally, there are different expactations on one
thing which seems to be the same to everyones eyes.

This is not surprising, there are so many expectations on the computer
idead it-self : is it a game machine ? a modeling machine ? a concret
tool ? an abstract tool ? Differents peoples use the same thing, the
computer, with different expactations.

After that is said, I can tell about my own expactation on
algorithms : I expect it to be something with can help me to see in
which direction I should go to get a good solution to a given request.
Something which would be abstract enough to temporaly avoid thinking
about real machines limitations, and which I will bring into a concret
methods in a concret langage, perhaps using other algorithms to solve
some requested steps to there.

The algorithm is also what I do in a reverse way, when I try to
understand why a particular piece of code does not work : I get rid of
implementation details, try to mentaly build an abstract algorithm on
which I will mentaly operate. By the way, I may or may not solve the
trouble this way : this view may tell me the algo is ok, and the
trouble is perhaps somewhere else (a browser bug, or a not well
understood context, or something else).

There are concret codes which applies very well to machine, and there
alternate view, more abstract, sometimes more fuzzy, with which
humans, who are not machines, deals.

(I hope I did not miss an important point)

> Vista is driving me mad. It's switching locale settings
> at unforeseen moments, makes me start tzping funnz words.
> Do thez ever use anzthing but Office inside MS?

Yes, I agree agree with this too :D (joking, with a serious though
behind)