Quotes of the week [LWN.net]

Usually my git problems are root-caused down to my lack of a PhD in hermeneutic metaphysiology, but not this time, methinks.

-- Andrew Morton

Kids: do not shove random modules into your kernel. Just because Linus does something doesn't make it a good idea...

We've moved half the kernel brains to userspace with udev, initrd and modules; it's really unfair that you're not sharing all that why-won't-my-machine-boot love.

-- Rusty Russell

[T]he kernel team has evolved from a small team of buddies to a large enterprise. And to survive this evolution, we may need to apply the immoral principles found in big companies.

-- Willy Tarreau

to post comments

Quotes of the week

Posted May 8, 2008 6:40 UTC (Thu) by madhatter (subscriber, #4665) [Link] (1 responses)

I always enjoy reading these quotes in context, and I'm grateful that you provide it.  In this
case, though, the Willy Tarreau posting linked to doesn't seem to contain the text quoted in
the QotW article.

Oops

Posted May 8, 2008 13:01 UTC (Thu) by corbet (editor, #1) [Link]

Hmm, don't know how that happened. Fixed now, sorry for the confusion.

Quotes of the week

Posted May 8, 2008 18:25 UTC (Thu) by jd (guest, #26381) [Link]

Usually my git problems are root-caused down to my lack of a PhD in hermeneutic metaphysiology, but not this time, methinks.

This has to be the best quote in a long time. If any University chancellor reads LWN, can you please give Andrew Morton an honorary doctorate in hermeneutic metaphysiology, so that he can continue working on the Linux kernel? Or maybe a Masters in paracomputology? Some award screwy enough that it doesn't harm the credibility of either giver or recipient, but rather serves to illustrate his substantial genuine contributions and also illustrate by the vehicle of humour the genuine difficulties faced not only by Linux but by the majority of University endeavours as well of maintaining coherency across very large scale multi-disciplinary groups. It can do no harm, and it might even improve attitudes towards the management of such projects.

Quotes of the week

Posted May 8, 2008 18:37 UTC (Thu) by drag (guest, #31333) [Link] (3 responses)

> [T]he kernel team has evolved from a small team of buddies to a large enterprise. And to
survive this evolution, we may need to apply the immoral principles found in big companies.

*BTTZ* WRONG. Try again.

Now I am a big pro-capitalist guy and I accept that it's big corporations that make the world
go around (while small ones do all the work), but there are some things that should be avoided
that are common to large industries...

American-style management practices. These things are f-ing horrible. The very idea of
establishing a sort of adversarial competition internally (or even externally) in order to
help manage things is just one very huge hunk of _very_bad_idea_.

This management school of thought says such things as employee competition will lead to
improvements and higher rates of output. People are encouraged to do good and discouraged from
doing bad.... This totally back-fires almost every time.

Industry works because the workers are so good at what they do. If they are able to do good
work it's _DESPITE_ those style of management practices, not because of them.

You see...
People _ALREADY_ want to do good work. They don't want to do bad work. They don't want to
introduce bugs or let other people deal with their dirty laundry. They introduce bugs into the
code because that's just what human beings do. You can play games and try to force people's
hands at fixing bugs by refusing to accept other hunks code and all that.. but it really isn't
going to work except under specific cases. (this means that occasionally, yes, refusing
patches are called for, it's just the exception that makes the rule)

Everybody here involved in kernel development is involved because they _want_ to do good work.
They are all very smart people and probably can do any number of things that will make them
far more money. There are tons of other kernels and other peices of software they can be
having fun on. You want them doing what they are good at.

Purposely introducing inefficiences into your system of development in order to force some
sort of positive outcome is just bad. The answer is not to slow down development... you _want_
fast, efficient, development. This is a good thing.

Bugs are going to happen. There is nothing you can do about. Big changes bring lots of small
bugs with it. Drivers that don't get used very often, or are only used under certain
circumstances will have more bugs.

Right now the Linux development stuff depends entirely on it's developer base to detect bugs
and absolutely depends on the end user to report bugs. In fact those end users are as
important as anybody else in the software development process.

Ok... as programmers you guys love to generate lots and lots of data.

What is needed is some solid statistical analysis of the Linux development process. This is
the _GOOD_ stuff from American industry. The old-school stuff of statistical analysis that
certain American industrialists* developed in the 40's and the Japanese adopted in the
50-60's. (and the Americans quickly forgot leading to the low-quality industry we had in the
70's)

Identifying and eliminating weaknesses in the design and process, not trying to force workers
to do something they can't. It's about cooperation towards a common goal, not finding some
quick-fix trick to make people work harder or be forced to do what you want.

Something like... driver bugs are unavoidable. Obscure hardware and lesser used hardware and
buggy hardware will all introduce more bugs no matter what you can do... even if the
programmers are perfect (which they are not). So what can you do minimize the effect of these
bugs since they _cannot_ be eliminated in any practical sense?

*see: http://en.wikipedia.org/wiki/W._Edwards_Deming

American management practices vs kernel development

Posted May 9, 2008 17:15 UTC (Fri) by giraffedata (guest, #1954) [Link] (2 responses)

Now I am a big pro-capitalist guy ... Industry works because the workers are so good at what they do.

Looks like a pretty abrupt switch from Smith to Marx. A capitalist would say industry works because the people with the money spend it so wisely.

In particular, my impression of the internal competition concept isn't that it encourages people who would otherwise do poor work to do good work (which I agree is silly) so much as that it separates the naturally good workers from the naturally poor and eliminates the latter (from this work).

I wonder how you address the fact that the free market apparently selects these horrible American management practices -- i.e. companies that use them apparently give people more of what they want than their competitors who don't, resulting in the former having the money to continue and the latter going out of business. (I assume the management practices you mean to refer to are those that are the norm in the largest corporations)

American management practices vs kernel development

Posted May 9, 2008 20:22 UTC (Fri) by joern (guest, #22392) [Link] (1 responses)

> In particular, my impression of the internal competition concept isn't that it encourages
people who would otherwise do poor work to do good work (which I agree is silly) so much as
that it separates the naturally good workers from the naturally poor and eliminates the latter
(from this work).

That is bull****.  A better way to look at a market economy is a question that came up in the
1990's changes in eastern europe and the former soviet union: "Who decides how much steel to
produce?"

After you laughed for a bit, you might realize the deeper meaning.  The main point of a market
economy is to 1. distribute the decision process and 2. award those who make the best
decisions.  Nothing is needed to achieve 2, it always follows naturally once you have 1.

If it was about good workers and poor workers, how can you explain the dramatic economic
improvements in countries going from command economies to market economies.  Massive layoffs
only work in companies within larger countries, because the created externalities are paid for
by others.  China couldn't get away with such a strategy.

American management practices vs kernel development

Posted May 10, 2008 2:29 UTC (Sat) by giraffedata (guest, #1954) [Link]

The main point of a market economy is to 1. distribute the decision process and 2. award those who make the best decisions. Nothing is needed to achieve 2, it always follows naturally once you have 1.

I can't tell what you disagree with, though you called something bull****. The points above are entirely consistent with the concept I described of internal competition generating benefits by separating good workers from poor workers. The decisions that are being distributed, and rewarded, are primarily the decisions on whose work product to buy.

If it was about good workers and poor workers, how can you explain the dramatic\ economic improvements in countries going from command economies to market economies.

Poor workers are removed from the work they're (comparatively) poor at and selected for work they're good at. Command economies aren't very good at that, and it makes a big difference.

But I didn't actually say separating good workers from poor is what makes capitalism right for China. I said it's an argument for internal competition in businesses. In particular, I said it's more common than the argument that internal competition makes poor workers good.