Hello, 

I understand that a lot of effort has been pushed in making 
Linux 3.3 the trunk kernel, and I understand that I probably 
missed long (IRC?) discussions on this very subject, but since 
3.3.8 is going to be the last supported kernel in the 3.3.x 
branch it might be a good idea to move on to another, more 
recent kernel version - and to do it slightly better. Not 
that anything is really bad, but there were obviously better 
choices that 3.3 at the time it came up. 

First and foremost, 3.3 has never been a -longterm. In fact, 
3.(2n+1) are doomed to be supported only for a short period 
of time, given that:

 * they are not supported by any -rt patch
 * and thus, will probably not promoted to -longterm.

The only existing, official -longterm at this point is 3.0, 
as stated on GKH blog [1]. The very first argument against 3.3 
would have been that given that 3.0 was already known to be a 
-longterm at this point, it should have been selected. 3.0-rt
are also going to be maintained as -longterm by Steven Rostedt.

Ben Hutchings proposed to maintain a -longterm of Linux 3.2
when GKH dropped its support ; Steven immediately stated that 
3.2-rt are going to stay for a long time as well. 3.2 is going
to be used by Debian and Ubuntu for (quite possibly) a long 
time, so we can be sure that this version is indeed a longterm
although it's a kind-of community-maintained one. 

I know that these lines do not add much to the debate. I get
that. The thing is, OpenWRT is supposed to be used in, well, 
wireless routers - possibly commercial ones. Commercial consumer
products need to have at least two things : 

 * they need to be able to pick a recent kernel when they
   start the development of their product. There is no real
   point today to select a 4 years old kernel "because it's
   known to work". Given the development model of Linux, 
   kernel breakage are less and less frequent (they still
   happen, but then the regressions are spotted quite fast). 

 * the need to pick a kernel that they know will be supported
   for some time. This decrease the workload on the kernel
   crew in this company, and increase their productivity. 

A third point should not be left aside:

 * sometimes, having a realtime kernel is necessary. Some VoIP
   apps mandate this (there is a reason why RTP is called RTP).

These should be the points to consider when its time to select
a new kernel version to work on. It may mean that the OpenWRT
tree needs to store two different kernel (a known longterm - 
maybe - the current kernel version).

There are also important things to factor in. 

 * the -rt maintainers pledged to maintaina -rt patch for
   every 3.(2n) releases (and possibly for every 4.(2n) 
   releases after that). That means that 3.(2n+1) have 
   virtually no chance to become a -longterm since 
   distributions are more and more interested in proposing
   both the -rt and the !-rt kernel to their users (see
   Debian, for example). 

 * One -longterm will be selected every year and will be
   maintained by GKH. Given the development pace, it will
   probably happen around 3.6 since 3.0 has been elected
   last year. The selection of 3.2 by Ben Hutchings is
   not related to the selection of another longterm by
   GKH.

Anyway, whatever your (OWRT maintainers) choices are, we're
going to accept and support them. But it might be good to at
least publish some kind of rationale document when a new 
kernel is pushed in the tree.

Hoping I have not been too disruptive, best regard,

-- Emmanuel Deloget


[1] http://www.kroah.com/log/linux/stable-status-01-2012.html
[2] http://thread.gmane.org/gmane.linux.kernel/1284809/focus%3D1285786

_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/mailman/listinfo/openwrt-devel

Reply via email to