Re: pre egcs-1.1 testing and Linux 2.1.x

["David S. Miller" <davem at dm dot cobaltmicro dot com>]

   Date: Sun, 23 Aug 1998 17:10:55 -0600
   From: Jeffrey A Law <law@cygnus.com>

   But I think folks may be missing an important point.  Poorly
   conceived and designed features should be removed, or disabled.
   That goes for any hunk of software, whether its the compiler,
   kernel, libraries, etc.

One thing people often fail to realize, is that the kernel issue is
totally different.  In the kernel case, you MUST stop the feature on
the way in, because you can't take it out.

When you remove a feature from the compiler, people may have to change
some flags passed to it or some parts of their code the _next_ time
they wish to compile their sources.  On the kernel end, if you remove
a feature, _existing_ applications will stop working.

So the "poorly designed features should be removed/disabled" arguement
cannot hold for the kernel.  We have to cut off new features on the
way in, before anyone can conceivably begin to use it.  Because once
it goes in, it has potential users precisely at that moment.

If you don't agree with me on this point, consider the case where it
was globally decided that SYSV streams are ill conceived and should be
removed from any kernel which has it, then try to get Solaris to
remove streams from their kernel :-)

Later,
David S. Miller
davem@dm.cobaltmicro.com