Dr. Roy Schestowitz Latest posts | Real-time contact
Short bio: Software Engineer, interdisciplinary researcher, and an advocate of fair competition (read more)
Short bio: Computer Scientist, FOSS supporter (read more)
Tux Machines (TM)-specific
Submit search form
I have heard the topic discussed in various forums and points of view.
Which would you say is the better choice for a linux based server?
Please give reasoning for your answers and not post "sux" or "rules" nonsense.
For servers - Static release/repo.
The "theory" of rolling releases is great, but the real world application, not so much.
Servers MUST be stable and secure. With a rolling release, you rely too much on the upstream vendor not to fubar something your system must have (not that it can't be done - mainframes have been doing rolling upgrades for decades - it's just EXPENSIVE to do it right).
RHEL/CENTOS has the right business model. Forget the fluff (and or bleeding edge stuff), only put well tested software into their repo's, backport security as needed, and support the whole thing for 5 years (or longer for security patches)
Of course it doesn't really matter what method the upstream vendor uses, you still need to run a parallel test environment along side your production environment, and test everything (and I mean EVERYTHING) in the first before rolling it out on the second.
It's just easier (for me anyways) to plan your server environments (and their future) if you have static (but not the ridiculously short 6 month timeframe) releases.
Rolling release is good for one reason. You get the full security and bug fix updates as intended by upstream.
No amount of backporting fixes is enough to keep a system secure and bug free. It's as simple as that. If I backport fixes from kernel git tree to a stable kernel 2.6.2x release, I'm most likely going to miss a lot of fixes. Cherry picking fixes for popular bugs only isn't a solution and causes weakness in Static release distributions.
The only requirement for a rolling release to work is to keep the base system as simple as possible. Theoretically, no downstream patching should be done in packages such as glibc, gcc or kernel unless it is a patch waiting to be eventually merged in a future upstream release.
Going forward, Tux Machines does not have ads. Instead it relies on readers' support and is run as a public service.
Help Support TM
Wall of Appreciation
Follow the site via RSS/Twitter.
Part of Bytes Media
Content (where original) is available under CC-BY-SA, copyrighted by original author/s.