Help Please

Kenichi Okuyama okuyamak at dd.iij4u.or.jp
Mon May 21 09:55:37 GMT 2001


Dear Andrew,

Before going any far, Why is you mail having reply-to which tries to
force reply to reach only to you? I think you need to look into your
mail environment....

# I can't give you any more hint because your mailer does not
# seems to set X-Mailer header..


>>>>> "AT" == Andrew Tridgell <tridge at samba.org> writes:
AT> We run Samba under insure to catch memory leaks. If you know of a
AT> memory leak then point it out rather than just sending such crap to
AT> the mailing lists.

I did. Look at old Mail and patch lists.


AT> Kenichi, you continuously make annoying comments like this. When you
AT> send patches that make sense and that we agree wiith then we will
AT> integrate them. Instead you say things like "all development must stop
AT> for 6 months" which is completely insane.

Insane, you say. But is it?!

Adding new function on top of memory leaking source code is like
building castle on top of mud. Castle will simply sink down.

What we need is brushup and cleanup of current samba code, so that
we can add functionalities. So that we can tune performance.
So that we don't have global buffers which prevent from changing
interfaces. So that we can separate string treatment from simply
digging hand into array of char and scrunble, so that we can face on
Unicode ver3, where we have something like 
"e with two dots on top" described in two words, instead of one.


I think continuing current work without any re-design IS INSANE.
All the listed above is something I can say within line. We have
much more to come, which is based on lack of understanding how to
treat buffers correctly and speedly. The problem is spread all over
the code, and need FULL BRUSHUPS. Needs totally different data
structure and need to add some OO idea.

Now, who can work for such a thing while target is moving?! New code
is being added every day and making bugs locked up with each other.
Those code generates compile error.

Giving PATCH? NON-SENCE!!!
Patch will grow larger than original code. And when patch is done,
you'll be on 2.9.8 or somewhere. You're not understanding what
you're saying.


AT> We have committed over 3000 changes so far this year. Thats over 20
AT> per day. And you complain that we are slow??

Yes. You are slow. Extreamely slow. It's because you're setting
wrong priority.


You're setting priority on adding functionalities, not on fixing
memory leaks and all other fixes that do not give anything to
function list. Not on changing structure for Internationalization.

Ofcourse, you're accepting fix to "NEW features" to work.
But that's not what most of the people are looking for.

Most of the people are looking for stability. Most of the people are
looking for something that won't be sead something about from CERT.


You should stop development for several month(Now that you added
even more code, it's not just 6 month, we need more). Annouce that
you stopped. And ask for fix of basic structure. So that Samba can
re-start growing into faster speed, and Intenationalizations.

We don't need small fixes now, we don't need new functionalities.
What we need now is design. Then structures and algorithms, then
brushups which removes all the memory leaks and security holes.

When all of them are done, Lets focus on new functionalities again.
I know this is what you really want to do, for that's what I feel
joyful too.

BUT! We always have graduation test before summer vacation.
And we need to pass them first, or come to Japan, and live here.

# In Japan, we have graduation test on end of Feb. Here, you can
# enjoy your summer without being disturbed by graduation test.

If you're in Japan, You can face this kind of problem more easily.
Multi-byte world will give you more sensitiveness to design, which
leads you to robustness and securenes.
---- 
Kenichi Okuyama at Tokyo Research Lab. IBM-Japan, Co.
               @Samba Users Group in Japan




More information about the samba-technical mailing list