Re: Parenthesized Target Notation (was: Partial retraction)

Peter Gummer <[email protected]> wrote:
	And the prettiest irony of all is that ISE Eiffel -- sometimes  
	castigated on this list as "ECMA Eiffel" -- 

That is completely untrue.

The phrase "ECMA Eiffel" was introduced to this list by me
and it has reference *ONLY* to the language described in the
draft ECMA Eiffel standard.

It has ABSOLUTELY NO REFERENCE AT ALL to ISE Eiffel,
for the very simple reason that I do not know what is in ISE Eiffel.
I have never used ISE Eiffel, and the manuals I have for it long
pre-date ETL2.  I have never at any time "castigated" ISE Eiffel
and the suggestion that ISE Eiffel has been "castigated" under the
guise of ECMA Eiffel is as unjustified as it is offensive.

	[ISE Eiffel] has had both 'clone' and  
	'twin' for a while now and, judging by its track record, will  
	probably retain 'clone' for some years yet.

BUt *IN A MODE WHICH CONFORMS TO THE ECMA STANDARD* it *cannot* allow
'clone' or 'equal' because their interfaces use a feature which is
explicitly NOT allowed in ECMA Eiffel.

	Which just goes to show  
	that, although the backward incompatibilities in the standard may be  
	of deep academic interest, in practice Eiffel compilers don't have to  
	break existing code at all.

They can EITHER support old code
OR they can conform to the new standard
BUT THEY CANNOT DO BOTH AT THE SAME TIME.

As I have been at some pains to explain before, with respect to
assignment attempt, it is precisely guaranteed for both *in the same file*
*in the same compilation* which is required for smooth transition.

I am quite prepared to believe that ISE Eiffel *will* continue to support
things that it previously supported for a long time, but either that is a
different compilation mode from the one that supports the ECMA Eiffel
language or it has to be rather more forgiving of standards violation
than people writing to a new standard would be happy with.

I note that Sun's Java compiler has a command line switch saying which
version of the language to use.  That's the kind of thing I'm talking
about.  You can compile a .java file *using the same compiler* for
*either* Java 1.2 *or* Java 1.5, but you cannot compile it for both
*at the same time*.  The parallel should be obvious.