Re: away

在 Thursday 30 March 2006 18:55，Dag-Erling Smørgrav 写道：
> 
> David Xu <davidxu_at_freebsd.org> writes:
> > I will go away today due to some unpleasant person attack to me, in
> > the past, I have made lots of work in FreeBSD threading work, this
> > includes kernel threading in earlier stage and thread libraries
> > later, and gdb support for these new thread libraries from kernel to
> > userland everywhere, spent lots of time to work in libpthread, and
> > later make libthr to be best performance library for mysql and
> > possible other applications developed on Linux and make it run on
> > most platforms we current support.
> 
> I've been reading and re-reading the relevant threads on -threads,
> -current and -developers.  The more I read, the less I understand.
> 
> At the tail end of two days spent immersed in libthr, I sniped at you
> on -committers.  That was wrong.  I understand that you were insulted.
> I compounded my mistake by not apologizing immediately.  I'm not
> trying to defend that.
> 
> To spell it out plainly, for the record: David, what I wrote to you on
> -committers was unfair and unprofessional, and I apologize.
> 
> However, I do not understand how things developed from there.  I tried
> to start a technical discussion with you on -threads, explaining what
> I thought was wrong with libthr, hoping that you would understand my
> concerns, and hoping to work with you to correct and improve the code.
> This did not work out.  Perhaps you had already made up your mind to
> leave; perhaps you were too hurt to read anything but insult into
> anything I wrote.
> 
I didn't have idea to leave the project just before your rush, I was
discussing with deischen_at_ about versioning things, and libc bump etcs.
and I wanted to get an approve from re_at_ to merge rtld locking from -HEAD
to RELENG_6 for libthr, I was still improving libthr, that's not even a final 
shape, e.g, I had planned to implement C++ object destruction safe with
pthread_exit(), and implement a structure based not pointer based
pthread synchronization objects, can be shared between processes,
current FreeBSD does not support these, but other systems do!

of course, I am not satisfied with current umtx interface, I do know, I even
want to recreate a whole new interfaces to make things more clean,
I talked about lwp mutex which supports robustness and priority propagating,
the current umtx will be outaged, deischen_at_ should know this, I had
a private communication with him.

but your words really hurt me, you meant I should go back and study
CS lesson again ? this is definitly an insult.
I graduated from Computer Science in college in 1993 summer,  my first
programming language is Pascal learned in 1989, second is C, and third
is X86 assemble language, I studied how to operate a computer on VAX 8200
system, I still remember its green character on terminal and longer command
line than unix did, I also use unix system which to learn shell script,
I learned programming on PC, there is Apple II in lab, but I dislike it, PC is 
8088 based, only has 640K memory, but can run turbo C and Borland C++ etcs,
I wrote some very funny code, TSR, TUI (better than current
ugly freebsd installer!) my graduation task is to mannually setup a micro 
processor using very simple chips and hundred of  physical lines, and use
switch to input micro code and data and strobe it into the chips, and run the 
program and print data on LED.
I learned how to programming in protect mode by 32 bit dos emulator,
figure out how to implement a GUI system, rectangle clipping, layer and focus
management, 2D graphics clipping, etcs, the code is still on my old CD.
I touched Linux at 1997, used to be a system manager and a programmer,
needn't to day, linux's ealier version has serious performance problem under 
load, so I went to FreeBSD, and met julian on net, started work on KSE.
The KSE work was started from zero, nobody knows how to make a full threaded
system, it was a very hard time, there was debate and rush, this made us
very bad, I don't want to say more. I picked up libthr and a rotted umtx 
interface, try to make other peoples to believe that M:N might be not a right
way to go, as I know it is very diffcult to make it SMP friendly, it's my
personal view, because I tried to implement a SMP scheduler and got the
feeling, I also fixed some stability bugs in ULE scheduler. I may be totally 
wrong.

> I am not trying to rush you, as you implied, and I would not commit my
> patch without your approval; if I did, I'd be out on my arse faster
> than you can say "backout war", and with good reason.
> 
> My initial impulse was to tell you to grow a thicker skin.  I'm not
> going to say that, because it's stupid and insensitive.  What I'm
> going to say, instead, is this: I am only one of several hundred
> FreeBSD developers.  I have no special position or authority in the
> project.  My opinion of your work is not representative of that of
> other committers or of the project as a whole.  What's more, while I
> may disagree with some of the choices you made in the course of your
> work, I do not disapprove of your work in general.  You should not
> view my criticism of the umtx interface as more than what it is:
> criticism of one small part of your work from one single developer.
> 
> In closing, I'd like to ask you to reconsider your decision to leave
> the project.  I don't think it's in anybody's interest to have you
> leave; and whatever you may think of me, it was certainly not what I
> intended when I started looking at libthr.
> 
> DES
> -- 
> Dag-Erling Smørgrav - des_at_des.no
> 
> 
see above, I am in a bad mood, I don't know when I will return, wish you can
work better without me, I beg pardon if some mistakes I  made and some
words I said here brought inconvince to you and all other people.

Best wishes to all FreeBSD people.

David Xu