Re: RedHat users' BEWARE !!

["Madhu M. Kurup" <madhu@xxxxxxxxxxx>]

Hello,

	Before things denigrate into a flame war, I hope that people can read up on many components of the problem:

Raghavendra Bhat mailed to LIG with:
>Hello:  Red Hat users' migrating to Red Hat v7.0 are 
>warned of severe incompatibilities.  Please read the
>below given link:
>
>http://gcc.gnu.org/ml/gcc-announce/2000/msg00003.html
>
>This shows that release cycles have to be brought down.
>There is no point is bringing out NEW versions if it is
>going to be totally broken and incompatible.
 
First, take everything you read with an elegant pinch of salt. What the comittee wants to say is the following:

2.95.2 is the official last release
2.96   is the RH released devel snapshot
2.97   is now the same compiler, except that now the GCC committee doesn't want to touch the RH snapshot with a bargepole.

	Warning of "severe" incompatibilities is like warning people from not using 2.4.0-test? just cause it is not a release version, but in devel. As a C++ programmer, some of the frustruation that I feel with the C++ support in 2.95.2 is clearly removed with 2.96. And at the bottom line that's what I need, a compiler that will compile ISO Standard C++. 

	What people fail to realize that 2.95.2 is already totally broken and incompatible (with the ISO standard at least, especially in the stream classes, if you are so interested!). 

	For me what is so refreshing in Linux is that it has no reverence for tradition or established practise. Linus is all too capable of taking years of tradition, evaluating it and throwing it out of the window. I was recently reading a thread on lkml (via kernel traffic - http://kt.linuxcare.com/kernel-traffic/latest.epl ) where he said that Linux threads would not have POSIX semantics simply because they were dumb / inefficient / broken / whatever. A certain gentelman (who was on the POSIX committee, IIRC) then stood up and complained. :-)))). The point is this : if something works and works well - it should be used. If it's efficient and useful it should be used. Not because it's recommened and "right".

	Following theory and committee's may be fine. But I wonder if they work. Consider the following examples that come to the mind - the ISO-OSI network layer (only in textbooks!), glibc (adopted by RH in 5.0 when everyone else complained), KDE (Re: Debian non free, blah blah). 

	Linux is all about choice. So, you don't want a release compiler. Don't install RH7.0. Is anyone forcing you to? If you'd want your stream C++ code to compile atleast then you would!

Cheerio,
Madhu

   Madhu M Kurup /* Nemo Me Impune Lacessit */ madhu@xxxxxxxxxxx