Re: Air traffic control system in Java

[Georg Bauhaus <rm-host.bauhaus@maps.futureapps.de>]

On 3/3/11 11:40 PM, Hyman Rosen wrote:
> On 3/3/2011 5:36 PM, Georg Bauhaus wrote:
>> Found this:
>> http://www.atego.com/products/aonix-perc-raven/
>> But I think using Java in this case means to use @AnnoTations
>> extensively. They will extend Java to be a language better suited.
>
> I'm not sure if you're being sarcastic or not. But just in case,
> remember ANNA: <http://books.google.com/books?id=tybCNcTXmUkC>

Yes.  (I hope that the aspect specifications of Ada 2012
will be better integrated with the language (and not require
any source transformation); certainly this should be the case
for Pre, Post, Invariant, and expression functions?)

Considering hard real-time support in a language definition,
can we not have a language for expressing the recurring
patterns that embedded systems programmers have been expressing
for years, but, apparently, using means outside the
language used?  Means like being careful, being smart,
being knowledgeable, being proud. Then there are tools that
flexibly mimic for C what is built into other languages.

Do we have to fool ourselves into thinking that just switching
languages is out of the question?  What makes us believe
that it is easier to integrate two of them, Java and
annotations---though not actually creating something
really integrated from the start?  (I guess that's because
---whichever is the technically flawed language---
we'd have to admit that we have made a mistake by choosing
the language and fear the uncertainty of the new or that we
will have to give up legacy libraries, or will have to prove certain
algorithms again.  And give up the language we have
chosen aloud.  OTOH, starting from scratch is considered
a healthy approach, too.  I find this contradictory
from a technical point of view. Is it economically rational,
balanced, impartial?)

There is knowledge and experience.  Why not condense them
into one formalism?  Mathematics has little trouble with
common formalism.  Any differences in mode of expression don't
change the subject matter.

Looking at special databases for long series of data like stock
exchange prices, real-time or past-time (kdb by Kx); a special
dialect of APL (k) serves to compute results from those data.
In a niche of customers who would use such a system it
seemed a good investment; the language is specially made for
the purpose.   And the thing is being developed as needed, IIUC.
(I understand that Sybase has something on offer, too.)

But in comparison, real-time systems in general do not really
fit a niche, do they?  Far too many systems, far to many
programmers for a niche.  And lots of semi-formal knowledge
regarding proper use of (mostly) C.  (Somewhat less of "semi"
if Ada is used, since Ada formalizes some of the semi-formal
knowledge (as was required originally).)

Why doesn't industry sit down for a moment and consider what
kinds of real-time algorithms they still want to be able to
write and then ask for that language?

Some days I am dreaming of market pressure exercised by every
single customer: we won't pay any compiler maker for two years.
Instead, we will spend 100% on thinking about what we want.