Re: Parenthesized Target Notation (was: Partial retraction)

"Peter Gummer" <[email protected]> wrote:
	Ok, Richard, a fair argument, although we may find
	counter-examples in history (C99 perhaps?)  :-)
	
The main example I've actually been using is Fortran 95.

I would say, yes, C99 is *not* "just a new version of C89",
precisely because in the general case migrating code from
C89 to C99 is *not* just a matter of recompiling.

However, the C89 standard *warned* that certain things were going
away, and if memory serves me correctly, C99 hasn't deleted anything
that we weren't given 10 years notice of.  So there was a 10 year
period in which C compilers were *required* to support both the old
features that were going away and the new features that replaced them,
so that it was possible to slowly and steadily convert while keeping
your code both correct and portable.

In the same way, Fortran 95 dropped some obsolete features of Fortran,
some of which were warned about in F90 (but which were widely advertised
for a couple of years before F90 actually came out) and some of which
were warned about in F77 (so people had had 28 years to upgrade, which
should be enough).

If there were a 'diarthrognathus' ECMA Eiffel 2005 and then a revised
ECMA Eiffel 2010 (or even ECMA Eiffel 2007) which dropped the old
features, no worries, as long as there actually were a few Eiffel 2005
compilers around.

	But I was actually replying to one of the SmartEiffel guys, who
	seem to be adopting the position that they are Eiffel and ECMA
	is not.  According to your reasoning, Richard, SmartEiffel
	ceased to be Eiffel as of version 2.0.

Er, yes.  SmartEiffel is a member of the Eiffel "family", just as Pascal
is a member of the Algol "family".  In fact when I found out how much
backward compatibility had been lost I was very upset.
	
This is one reason why I was really hoping for an "industrial quality"
ECMA Eiffel standard, which would provide a gentle transition, and to
which I could appeal for support against too much loss in SmartEiffel.

This much I will say for the SmartEiffel team:  they don't owe me anything.
They are giving their stuff away free.  I don't _have_ to use their system.
I don't _have_ to upgrade from SmartEiffel 1.1.  Their primary job is to do
research in how to implement OO languages and systems, and if their software
is useful to anyone else, that's basically good luck.

The problem is that languages that are _like_ Eiffel but different have a
habit of dropping out of use.  Anyone still using Sather?  (And iteration
in Sather looked really cool.)  Anyone even _heard_ of Tethys?