Divorce the 3 tiers [was Re: My resignation]

Hui Zhou zhouhui at wam.umd.edu
Tue Jul 13 08:32:42 PDT 2004

<hijacking the thread here, sorry>

I believed the 3-tier development was the cure for the lfs crisis we 
are facing. However, it never developed into a structure that was in 
my mind originally. Currently, the 3-tier is related too much: 
Experiment in unstable, THEN testing in testing, THEN publish in 
stable. Despite the wordings from the editors, these 3 tiers working 
as in the last sentence can't accomodate different needs and opinions, 
both from the editors and the audience. As demonstrated in recent a 
few months, arguments never have stopped, and people with low 
tolerance kept suffering. As I view it, the 3-tier structure as is has 
no difference from the one tier development before. Correct me if I am 
wrong, all editors are focused on the unstable, just as before that 
every one is focused on CVS head.

How about divorcing the 3-tiers (which was in my mind all along)? 
Ustable and testing and stable have very different goals. Unstable is 
for Be at Edge, and can't tolerate lengthy discussing before 
committing; Testing is more like the book as before (when Gerard was 
active), through discussion before commitment; Stable is for 
publishing, targeted for paper books. Each tier can't work together as 
the clashes are doomed as we witnessed so far. Each tier should have a 
seperate group of people working on it, or at least have one 
authoritive person to lead. As a side note, I don't think we have too 
few contributors. Two or three core developers are really needed for 
each tiers. For example, Belfs was given birth by three people(Correct 
me) and if these three people are real authorities to maintain it, it 
will be more healthy for both book and people.

It's a pity that nobody have worked on testing after belfs was 
included. The testing branch should continue to be the active 
developement branch as before, have discussions like before. Testing 
tier should select packages INDEPENDENTLY, although now when 
considering new packages, there is a good reference -- the UNSTABLE. 

The stable is the only one that need version tags, for publishing. 
Maybe Gerard can still actively lead this branch? 
With this model, the unpleasant argument about udev/hotplug and other 
similar thing really shouldn't happen. Weather to include in the 
unstable is up to Jeremy and other core developers in UNSTABLE branch. 
Weather to include in testing is upto mailinglist discussions as 
before and with the final say of the active leader. The core 
developers in unstable should not be upset by whatever discussion come 
to  because it really won't affect the unstable.

The seperated 3 tiers is not only necessary to value the effors and 
talents or developers, but also necessary to satisfier different group 
of audience. There are numerous people who only monitor the unstable, 
and there are quite a portion of people use testing. From time to 
time, it is possible the unstable may be more stable than testing, 
there will be nothing wrong with it but an interesting thing to see.

Best Regards

Hui Zhou

More information about the lfs-dev mailing list