Linux kernel hackers at O'Reilly Open Source convention

by Andy Oram

Related link:

The second annual Portland kernel hackers' BOF took place last night,
bringing in five men and two women programmers to speak to a dozen
serious, hard-core Linux enthusiasts. Portland is a major, if scantly
appreciated, computing site, home to a large number of Linux kernel
developers (mostly working on Linux drivers and driver-related
subsystems) who meet socially once a month and are employed by OSDL,
IBM, Intel, and a variety of other local companies.

It was a pretty self-assured and convivial bunch, offering such
opinions as:

  • BitKeeper, a proprietary version control system used by the Linux
    project (and, incidentally, MySQL AB) is a necessity (although a
    couple of the hackers disagree) that is "tailor-made for what we do"
    and not yet approached in features or stability by any free software.

  • What's good for the desktop is good for the server, and vice
    versa. Desktop users are anxious to have faster boots, for instance,
    but so are server administrators. The improvements in the kernel made for
    one type of system enhances, rather than detracts from, the kernel's
    appropriateness for the other type of system.

  • On the other hand, many things that impact the performance of
    GNU/Linux systems lie outside the kernel and the kernel developers'
    control. For instance, version 3.x of the gcc compiler produces
    much larger executable code than version 2.x, so version 2.x is
    currently recommended. As for user-space systems such as GNOME, kernel
    developer Greg Kroah-Hartman says, "we regard themas load tests."

  • Do we need real-time Linux? Linux is already pretty darn good for
    real-time. Very few projects need hard-real time. Some developers have
    been persuaded to use vanilla Linux for real-time projects and to
    switch when they discover a need to--and have not yet discovered a
    need to.

Copyright issues raised their head, too, as one would expect. (A Moot
Court on Monday night, led by professor Pamela Samuelson, laid out
both sides of the SCO v. IBM case, not necessarily with equal respect
for each side, and led to a lively discussion of the legal risks in
free software development.) The kernel developers agree that one
should ask contributors to verify that they have the right to donate
their code, but think that the safeguards put in place by major
contributors (notably IBM) are more than sufficient to ensure the code
is clean.


2004-07-28 09:58:47
For non english speaker
What do it mean :

we regard themas load tests

exactly in english ?


2004-07-28 10:05:34
For non english speaker
"themas" should read "them as".

To clarify further, the (probably humorous) remark means that the desktop environments suck up enough processor time that they can be seen as a good way to test the scheduling and other kernel algorithms when the system is under heavy stress.

2004-07-28 23:11:51
For non english speaker
it means they don't take real application seriously, see them only as a means to find out the performance issues resulting from running CPU or RAM intensive programs.