[Oberon] Appeal for greater formalism vs. quick temporary fix.

[eas-lab at absamail.co.za]

My re-reading the following post emphasised again the need for
greater formalism:-

Stefan Salewski wrote:-
> when rounding Reals to Integers we have at least four ways:
> 1. rounding to -Infinite (floor() in C)
> 2. roundind to +Infinite (ceil() in C)
> 3. rounding to nearest
> 4. rounding to zero
> 
> ENTIER(x) will provide rounding to -Infinite,
> ENTIER(x+0.5D0) should provide rounding to nearest,
> -ENTIER(-x) should provide rounding to +Infinite.
> 
> IF x>0 THEN RETURN ENTIER(x) ELSE RETURN -ENTIER(-x) END;
> should provide rounding to zero.
> 
> Is there a better (faster) solution.
> Sometimes I use x:=ENTIER(x+0.5D0).
> I guess this is not very fast.
> 
> In http://www.oberon.ethz.ch/native/WebAlpha.html I read:
> 
> > Reals.Mod - Default rounding mode is now towards negative      
> > infinity (like ENTIER). This allows ENTIER to be compiled much 
> > more efficiently, because it does not need to set the rounding 
> > mode. If you set another rounding mode with Reals.SetFCR, be   
> > aware that ENTIER will also use this rounding mode! 
> 
> I do not fully understand this. Which parameters does 
> Reals.SetFCR() need, which other functions exect ENTIER() does 
> Reals.SetFCR() modify.
> 

This seems also to relate to the recent 'DIV query' ?
It seems to me that there is a heirarchy of authorities, which must be,
obeyed.   Because 'just hacking it' is such fun, a formalism needs to be
in place (like a constitution) to impose discipline.

I'm guessing (could be very wrong) that the numerical roundings
are defined by 'the matematical authorities', and n-o's task is to
implement those higher definitions. 

> > compiled much more efficiently

should be replace by 'overall more efficiently', where the user's
time/effort is the overwhelminly major component of overall efficiency.

As a recent contributor observed: saving machine cycles seldom
makes sense today if it wastes users time -- overall.
========================

Working formally from the 'higher authority' specification is 
vital for internet usage.  

> > I could investigate the RFCs, but am reluctant to do so, given my 
> > perception of ETH's 'management policy': individuals hacking away
> > privately and submitting results without feedback/openness.
> 
> Edgar wrote:
> As long as it concerns PPP I would be happy to receive bugfixes or
> improvements.

OK good, but who is concerned with TextMail.Mod ?
There's much more than PPP to be looked into.
eg. TextMail.Directory expects  the number of mails on the server
to be supplied, which my new ISP doesn't do.   And when other
n-o users find they can't connect because of ISP changes, they will
just revert to M$.   Then everybody can be a top-poster.
============

 Previous poster(s) have mentioned n-o 'falling apart' as the
specialists become no longer available.    This is the problem 
with systems which are personality-based instead of 
organisationaly-based.  It's like a society: the colaborators
need to know what the rules are - independant of any one/few
person(s).


-- Chris Glur.