roger browne

Visual Eiffel 5.0 for Windows is now available.

See http://www.visual-eiffel.com/

Regards,
Roger Browne

roger browne

Please stop worrying. I'm not the only one who you have asked to read
the article again, and I think we do "deeply understand".

We just use the term "contravariant" as a convenient label for the
(possibly surprising) reversed conformance in the proposed new OpenRule:

"ROUTINE[B1, O1] conforms to ROUTINE[B2, O2] if
...
O2 conforms to O1"

What would REALLY help is if you or someone else can confirm the wording
of the new ECMA rule. I have been led to believe that it is now going to
be something like this:

"ROUTINE[B1, O1] conforms to ROUTINE[B2, O2] if
...
O2 and O1 are the same"

Is that correct?

Regards,
Roger Browne

jim cochrane

I assume that "people from that list" includes some ISE people. In that
case, I think you're on top of the situation (did I get that colloquialism
right?) - you've already taken care of the essence of my suggestion.
Please disregard my suggestion.

Thanks, Roger, for the update and for your diligent efforts WRT these
issues.

--
Jim Cochrane; jtc@dimensional.com
[When responding by email, include the term non-spam in the subject line to
get through my spam filter.]

roger browne

OK, I think I see what is going on here. You appear to be using the
unadorned terms "covariance/contravariance" to refer only to one
specific kind of covariance/contravariance (presumably
covariant/contravariant type redefinition for feature arguments).

But the word "contravariant" can be used to describe ANY two things that
vary in opposite directions. In this case it's the conformance rules for
(1) objects, and
(2) open agent arguments
that vary in opposite directions (with respect to the inheritance of the
underlying types).

Anyway, in the interests of clarity and consistency I'm happy to stop
using the word "contravariant" in this case.


Your explanation is a very good short summary of the paper.

Assimilating new concepts takes time. Catcalls were clearly exposed
for years before they were well-understood.


Your rule seems sound. If the ECMA committee say that they have found a
problem with your scheme then I'd sure be interested to see the
counterexample that they have found. Maybe it breaks in conjunction with
a catcall or double-dispatch or something similar.

I suppose I need to keep checking ETL3 revisions in case something
relevant pops up. I'd sure prefer it if the ECMA minutes were published.

Regards,
Roger Browne

miguel moquillon

Le Wed, 08 Dec 2004 18:52:35 +0100, Peter van Rooijen a écrit*:

No. he isn't mistaken. There is no a well-defined definition of the term
'variance' because languages are not scientific (thus not rational) and
their meaning depends mainly on context.
In our case, he gives us what the term 'variance' means in this context
and, in the light of this definition, what he said is correct.
If you are'nt agree with him, you have to be in the same context and not
in another one. There is not to say you are wrong ... in the case you give
your own meaning of the term 'variant', that is to say you carry us the
context on which you're building your argumentation. What you don't does
in this post...
More, the definition of 'variance' he gives matchs the one provided by the
OO paradigm of the computer science domain (perhaps not alone).


We have to be argee with the mean of each specific term we use in this
discussion before going on. After, this discussion will be more 'positive'
with a truely dialog than several monologs. Ununderstanding comes mainly
from different meaning of same terms by interlocutors.

Cheers

Miguel

georg bauhaus

:>> Variance is related to type variation in feature redefinition (this mean
:>> in sub-type).
:>
:> This is where you are mistaken. The term 'variance' means much more than
:> the narrow meaning you ascribe to it. Roger said essentially the same
:> thing, so perhaps you will entertain the possibility that it might in
:> fact be so.
: No. he isn't mistaken. There is no a well-defined definition of the term
: 'variance' because languages are not scientific (thus not rational)

Why not just say

{non,contra,co}variant WRT xyz

or similar when "xyz" does matter, obviously?
And let Humpty Dumpty talk elsewhere.

Languages, in various meanings of the word, are and have been the
subject of canonical sciencific disciplines.
To say that science implies rationality needs conviction, let alone
a single definition of rationality, which doesn't exist when it comes
to science.
Note that a *contradiction* is at the heart of computer science,
together with the impossibility to logically decide Church's thesis.

-- Georg

miguel moquillon

Le Thu, 09 Dec 2004 11:44:20 +0000, Georg Bauhaus a écrit*:

In fact no. Theses disciplines are 'humans' one, not scientific.
The term 'scientific' were applied to them in the 19 century with the
industrialisation way besause of the scientific displines success (math,
physics, ...). Lot of people tried to convince their 'human' displines are
scientific because they use like-scientific method or terms but in
reality it is just a coat.
In 21 century, we please to play with the definition of terms,
in another word to give us a different meaning that the word drives in fact.
Companies and politics play strongly with this game (see advertissements
for example).


See mathematical sciences for example.

Miguel

georg bauhaus

:> Languages, in various meanings of the word, are and have been the
:> subject of canonical sciencific disciplines.
: In fact no. Theses disciplines are 'humans' one, not scientific.
: The term 'scientific' were applied to them in the 19

In fact yes, these are sciences, what is a science if not human?
Superhuman? Mechanic? Divine?
The narrow notion of 'scientific', the popular prejudice you have
described, is only the hope or the arrogance of technicians mixing sound
epistemilogical issues with some superficial commonalities circulated
in the engineering communities. Mostly some superficial theory including
falsification. Popper himself has hinted to its limitations.


: See mathematical sciences for example.

The 20th century is the prime example, since mathematical
science has had to *define* what is allowed to be called "rational".
It turned out that "rationality" doesn't hold water, contrary to what top mathematicians had hoped.
See Frege -> Russel/Whitehead -> Hilbert -> Goedel.

But this is OT here, so I'll stop now.

-- Georg

cesar rabak

Lothar Scholz escreveu:

When I saw the (almost similar) post from Frieder and spent a time
browsing the site, I started to wonder in this line of reasoning...

I did not know there were so many ;-)


So now my optimistic way of seeing: once the project has been open
sourced, perhaps the best parts of both open source (perhaps there are
more I'm not aware of) projects can be put together?

mike meyer

Cesar Rabak <crabak@acm.org> writes:


There are five or six forks of SE at sourceforge alone. Most of them
are orphaned, though. I'm eagerly awaiting the "easy to build" source
release for Linux, so I can port it to FreeBSD.

<mike
--
Mike Meyer <mwm@mired.org> http://www.mired.org/home/mwm/
Independent WWW/Perforce/FreeBSD/Unix consultant, email for more information.