The sorry state of open source today

[Scott Long]

Dag-Erling Smørgrav wrote:
> The subject refers to an editorial by Radu-Cristian Fotescu, which was
> published on the author's own website and in The Jem Report:
> 
> http://beranger.org/feature/sorryfeature.php
> http://www.thejemreport.com/mambo/content/view/309/
> 
> The article contains several factual errors regarding FreeBSD.  I have
> posted a rebuttal on my blog:
> 
> http://maycontaintracesofbolts.blogspot.com/2007/04/sorry-state-of-jem-report.html
> 
> DES

I'll rebut you're rebuttal =-)

You're absolutely correct about feature-based vs time-based being a
problem.  However, KSE was NOT, I REPEAT NOT, the major nor the second
major reason for the FreeBSD 5.x problems.  5.x releases suffered from
the following problems that were much larger and much more immediate:

- ULE and the modularized scheduler
- PREEMPTION
- ATA
- UFS2
- Immature locking model, too much Giant

Now, I'll entertain that the KSE development caused hurt feelings among
some developers, but that was a professionalism issue, not a technical
issue.  I also do agree that M:N is a nice academic theory that has run
into real-world roadblocks, and that FreeBSD seems to be better off in
the end with 1:1 threads, just like most other OSes.  But KSE was a
stepping stone to get there; without it, who knows when we would have
moved passed libc_r?  It was a definitely a painful step, but it would
have been much more painful to not have any alternatives to libc_r.  I'm
glad that the project and certain developers in it had the courage to do
it AND to stick with it to resolve the tough problems.

Scott