People need to say "no"

Larry McVoy (
Thu, 25 Feb 1999 10:27:06 -0800

It looks like the pseudo-RT stuff is coming back which suggests that we
might have another lovely flamefest on the topic. I wanted to explain why
I repeatedly get involved in these obviously painful discussions.

I've watched several operating systems die. For the single reason that
nobody was paid to sit in judgement of every idea or line of code that
went into the kernel. It is far too easy to let the kernel grow in size,
or complexity, or in cache misses. It's easy - people show up and say
"I measured it, you can't tell the difference" or "it's just a little

People need to say no. In Linux, that burden falls on Linus' shoulders,
and we have come to depend on his good judgement and good taste. But that
is a long term losing strategy. What would we do if, God forbid, Linus
got hit by a truck? Or the more likely scenario where he just gets burned
out or bored with the kernel and moves on? Even if neither of these
happens, don't you think it is a little lame to count on Linus to always
clean up your mess?

Somebody needs to say no. All design and code needs to go through a
screening process and people need to justify their changes. If everyone
had Linus' sense of taste, his broad perspective on the whole OS, etc,
etc, then this process would not be needed. But Linus is remarkable,
I've never seen anyone quite like him in 15 years of OS work. Most people
are not like Linus, unfortunately.

People will get mad when other people say "no" or say "justify this" or
say "that's a bad approach". They will retort "who the hell are you to
tell me what I can do?"

The answer is that the no-sayers are nobody, it doesn't matter who they
are, it matters only that they guide people into doing good work. You
could actually produce most of the right results with an AI program that
just repeatedly asked "and why do you want to do that?"

The focus needs to be on the health and well being of the kernel.
Every change, every discussion, every action which affects the kernel,
they all need to be held up and discussed out in the open. You may
disagree, many people do, but the history of operating systems will show
these truths to be self evident, so to speak.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at