Quoting Florian Kaiser <mailinglists@florian-kaiser.net>:
>
> For anybody using Linux-Vserver in production mode, it is highly
> unlikely that they want (!) or would need the latest mainline kernel. No
> sane sysadmin upgrades to latest mainline every release. To maintain
> latest mainline, you also need a lot of resources (e.g. free time,
> testing maschines etc.) since they do change major parts of the needed
> subsystems very often. So by always providing patches for current
> mainline, you only support a small minority which probably doesnt use
> the kernel in production mode anyway.
This is not really true, the supporting cutting edge mainline is
essential for testing. This is how long term support releases happen,
we see what's good over a few years of testing and settle with it.
When I compiled the first 3.10 kernels I have someone say on IRC
something like "I'm not interested in a kernel so new for my servers".
For me that's not really an answer I want here. We need people testing
these kernels on as close to production systems as possible, that what
gives us the releases its good to support.
Further to that mainline is massively important for hardware support.
It's great to have a kernel which is stable and has a Vserver patch,
but if it doesn't run on your hardware or doesn't support a feature
you need, what are you going to do?
>
> Quite the opposite goes for LTS kernels. These only get bugfixes so they
> are almost hassle-free for most sysadmins to work with. I believe many
> who use Linux-Vserver use LTS versions. And from your perspective,
> keeping the patches working should also be a lot less effort. So by
> focusing on LTS, you provide support for the majority of users, while
> keeping your workload to a minimum.
>
> So, I suggest to put up some info on the wiki that from now on, you
> follow LTS lines and adopt supporting current mainline. That way nobody
> takes support for current mainline granted. And you could still work on
> it to keep up with changes, and as soon as there is a next LTS candidate
> in sight, you could then release a new LTS very soon after. All that,
> without the hassle nor the pressure to provide a tested "stable" patch
> for every major and minor mainline release.
Obviously its up to Herbert which kernels he maintains, but personally
I don't think LTS only would work. We need to put all the kernels
through the mincer, and see which ones are juicy. SFAIK that how the
mainline scheme works too. I'll be building 3.13.1 ASAP (I'm in the
middle of house move so not much time).
Cheers,
Ben
Received on Sat Feb 1 09:12:26 2014