[asterisk-dev] Proposal for DAHDI-trunk: deprecate old kernels

[Shaun Ruffell]

On Sat, Dec 31, 2011 at 02:40:45AM +0200, Oron Peled wrote:
> On Friday, 30 בDecember 2011 01:55:42 Tilghman Lesher wrote:
> > On Thu, Dec 29, 2011 at 5:34 PM, Oron Peled <oron.peled at xorcom.com> wrote:
> > > Yes. Since no one here has the same resources RH can put into
> > > maintaining their RHEL, this seems like the only resonable alternative.
> > 
> > Well, hang on a moment.  Since this does require Digium resources, we
> > will need their consent on whether or not they choose to support any
> > particular branch for an extended period.
> 
> Yes. The purpose of this discussion is to hear the thoughts of the
> different parties involved (AFAIK there are Digium people on this
> mailing list ;-)
> 
> [of course, also their unofficial opinions are most welcome as well]

Here is my unofficial opinion:

As long as the mainstream distros are maintaining branches for the
regular/production life cycle (7 years for RHEL afaik) in any
capacity, I'll patch at least trunk for those versions if possible.
This isn't to say I personally test with all the distros with all
the hardware (especially the older ones) but I'll try to resolve all
reported issues.

So the span of kernels that the trunk of DAHDI needs to support I
believe is:

[<oldest (possibly patched) kernel support by major Linux distributions>,
 <latest kernel.org released kernel>]

I try to do primary development on the latest kernels and on the
most recent "enterprise" distributions and then just test on the
older ones after the fact since most users who are setting up new
systems with new software are going to be using the newer
distributions typically.

Thankfully, 2.6.9 is going into the extended lifecycle here soon, so
2.6.16 will be the oldest kernel to support. Unless a valued
customer needs something different or a community member supplies
patches.

> > > * This means we normally have to maintain only a single LTS branch
> > > to bridge the final life-cycle of the previous RHEL version.
> > 
> > That's not true, though, if the extent of the LTS release is anything
> > less than the full 7 or 10 years that Red Hat supports their release.
> > Otherwise, we will have an unsupported LTS release near the end of the
> > Red Hat release cycle, which again puts IT managers in a very
> > uncomfortable situation.
> 
> 1. Expecting a single driver suite (DAHDI-trunk) to actually "support"
>    a 10 years span of kernel development (2.4-3.2) is....
>    "interesting" but not very realistic.
> 
> 2. Please note that even RH themselves have very limited support
>    after 5 years - Refreshed hardware only via virtualization
>      [https://access.redhat.com/support/policy/updates/errata]
> 
>    Do you really imply that Digium should offer better hardware
>    support for RHEL than RH themselves?

No, but I have not felt this has been much of a burden. Most
customers who aren't running on the maintained branches of major
distributions have been understanding if there were some hiccups...

> > > Last, but not least -- can we ring the bell on 2.6.8/9 (RHEL-4)
> > > immediatly after DAHDI-2.5.x is released, so we can remove it
> > > by February ? (no, I don't want to see this legacy in DAHDI-2.7)

I think it is fair that when RHEL-4 goes into the extended life
cycle the 2.6.9 specific bits can be removed from trunk.

> > I think you mean 2.6.  DAHDI 2.5 was released in August.  In any case,
> > we'd need for that kernel to be deprecated in a full release branch
> > before we remove it.  So, yes, it would need to remain in 2.7, in
> > deprecated mode, if it were not already deprecated in 2.6.  And we
> > would additionally need to declare DAHDI 2.6 to be an LTS release.  If
> > it were deprecated in 2.6, then 2.5 would need to be the LTS release.
> 
> 1. I actually meant after the comming DAHDI-2.5.x (0.3?), so that
>    DAHDI-2.6.y would carry the deprecation warning and we can clean
>    this code before DAHDI-2.7 release
>
> 2. The logic is because RHEL-4 ended "Production 2" phase almost a year
>    ago (meaning no support for new hardware) and would end "Production 3"
>    phase in two months (Feb-2012).
> 
> 3. While you are technically correct in your observation about the
>    need to declare some stable DAHDI branch as LTS. The real
>    question IMO is when to declare RHEL-4 *completely* unsupported?

Additionally I personally do not wish to declare any release of
DAHDI as Long Term Support (LTS) release. I don't think we need to
(yet?). I believe that the trunk of DAHDI should strive to support
all currently supported versions of Asterisk and on recent branches
as judgement dictates with the most common kernels. I think how long
to actively maintain old branches is best left a judgement call
based on the needs of the community/customers at any particular
time, with the goal to minimize the reasons anyone would not want to
or be able to run the latest DAHDI releases.

Again, I really wish there *NOT* to be any reason someone can not
run the latest version of DAHDI. The more people who are using the
latest, the better it is for everyone as opposed to having effort
divided up among multiple branches. There might be something in the
future that prevents this, but I haven't seen anything on the
horizon that would make this impossible.

> 
>      - If you think about the 10 years perioud (Feb-2015) than
>        by the same token why not support RHEL-3 (its 10 years
>        last until Jan-2014)
> 
>      - If you think about the 7 years (what RH calls production 1,2,3)
>        than that will finish in two months.
>        What LTS plan should we have for an OS that finish its 7 years
>        regular support period in two months?
> 
> [
>   RHEL-4 users, that's the exact place we should here you SHOUTING!
>   What? I can't here you? Anybody there? Using trunk drivers?
>   Great, because we realy need RHEL-4 testers for our newest drivers
> ]
> 
> > As I noted before, given that Digium maintains the repository and the
> > packages, this is their decision.  If they choose not to do LTS
> > releases for DAHDI, then I don't see any reasonable alternative other
> > than keeping support for 2.6.8 in all versions of DAHDI until after
> > the sunset date for RHEL-4.
> 
> IMO the sunset has passed about a year ago (5 years), we are now in its
> late evening and in two months it would reach the time of midnight kiss.
> 
> You are right that we/Digium can wait until its 3 more "nuclear winter"
> years pass. I argue that this is a bad decision.

I personally do not think there is a need to wait another three
years for trunk support.

Just my personal opinion,
Shaun

-- 
Shaun Ruffell
Digium, Inc. | Linux Kernel Developer
445 Jan Davis Drive NW - Huntsville, AL 35806 - USA
Check us out at: www.digium.com & www.asterisk.org