SEARCH

Enter your search query in the box above ^, or use the forum search tool.

You are not logged in.

#1 2013-04-08 16:52:58

Pierce Randall
New Member
Registered: 2013-04-02
Posts: 9

When is testing more stable than stable (if ever)?

I'm posting this here because it's not a support question.

I've been looking at the bug countdown for Wheezy, and I noticed that unstable has fewer release-critical bugs than stable.

Suppose that you wanted to use the most stable Debian base you could, where "stability" here means "free from bugs" (and not "never changes"). I would think that given that goal, the best strategy would just be to stay on Debian stable. However, you might think that the best time to switch to unstable would be the crossover point, when unstable has fewer bugs than stable. After all, there are fewer bugs, so wouldn't unstable be more stable?

What's wrong with this reasoning?

Offline

Be excellent to each other!

#2 2013-04-08 17:00:55

xaos52
The Good Doctor
From: Planet of the @s
Registered: 2011-06-24
Posts: 4,602

Re: When is testing more stable than stable (if ever)?

The hidden variable 'unknown bugs in unstable' makes it impossible  to answer your question.

If you have to maintain a server
then
      go for stable because you know it will have long term support and security patches
else
      whatever you prefer and can handle.

Last edited by xaos52 (2013-04-08 17:01:36)

Offline

#3 2013-04-08 17:02:16

Floi
#! Member
Registered: 2011-04-07
Posts: 63

Re: When is testing more stable than stable (if ever)?

I don't know all too much about the debian release cylce, but I think it might be true, as I recently read a post by Paul Tagliamonte, who concluded:

Hilariously, this makes me think I should be suggesting users to
switch from stable to testing when the “crossover” happens during the
release cycle.

big_smile


http://twitch.tv/jonfen - me streaming some noob starcraft: broodwar from #!

Offline

#4 2013-04-09 00:16:41

Pierce Randall
New Member
Registered: 2013-04-02
Posts: 9

Re: When is testing more stable than stable (if ever)?

Thanks for the link, Floi. Just what I was wondering.

xaos52 wrote:

The hidden variable 'unknown bugs in unstable' makes it impossible  to answer your question.

If you have to maintain a server
then
      go for stable because you know it will have long term support and security patches
else
      whatever you prefer and can handle.

Right. So one issue is the amount of unknown bugs. These aren't noticed on testing, and they someone finally notices them when the release is on stable, which explains the rise in bugs from stable. That would support the conclusion that stable is... stabler.

I wonder if another explanation for the rising bug count on the existing release includes incompatibilities introduced through backports or security patches. If that's a lot of the bugs, then stable will be less stable.

There's also the notion that bugs on stable will probably be less severe, since they took longer to notice and so probably do not come up in everyday use (if they're the first type of bug that I mentioned).

I'll keep this in mind. I plan on holding on to Waldorf for awhile after Wheezy goes to stable. But now I'm glad I went with it over Squeeze/Statler, since Wheezy was so late in the testing cycle when I got Crunchbang.

I prefer a relatively stable system not because I have a server, but because I want everything to work without crashing. I've come to appreciate having at least one stable system, which is why I'm glad that I'm on a Debian-based distro. (Well, one that isn't Ubuntu, at least.)

Offline

#5 2013-04-09 00:22:45

dura
Bloated Gimp
From: interzone
Registered: 2012-09-15
Posts: 2,106

Re: When is testing more stable than stable (if ever)?

Are you talking semantically or in practice?

Offline

#6 2013-04-09 00:44:32

VastOne
#! Ranger
From: #! Fringe Division
Registered: 2011-04-26
Posts: 10,163
Website

Re: When is testing more stable than stable (if ever)?

It is always intriguing to me, who runs SID all the time to see so much discussion about bugs in testing when I never see anything close to a crash in the supposedly 'unstable' side of Debian

And yes, I do run it on Servers.  When you know what you are doing and have the proper tools, it really is not a major science project to be successful


VSIDO | SolusOS

Words That Build Or Destroy

Offline

#7 2013-04-09 00:58:45

snowpine
#!-a-roo
Registered: 2008-11-24
Posts: 2,920

Re: When is testing more stable than stable (if ever)?

How can Debian Unstable, which is constantly changing and evolving, be considered "stable"? Here are Bruce Lee's thoughts: http://www.youtube.com/watch?v=On82lZQD8Bc


/hugged

Offline

#8 2013-04-09 01:05:38

VastOne
#! Ranger
From: #! Fringe Division
Registered: 2011-04-26
Posts: 10,163
Website

Re: When is testing more stable than stable (if ever)?

^ I never said it was stable...

I said it can be managed with the right tools and know how.  Making stable/unstable all in the eyes of the beholder

love the Bruce Lee though


VSIDO | SolusOS

Words That Build Or Destroy

Offline

#9 2013-04-09 05:01:55

DebianJoe
#! Code Whisperer
From: The Bleeding Edge
Registered: 2013-03-13
Posts: 1,207
Website

Re: When is testing more stable than stable (if ever)?

Debian's naming system causes a bias against the reality of volatility.  A much better system would be:
Stable=Enterprise/Corporate (when it costs you money when things don't "just werk.")
Wheezy=Personal use for Mom.
Sid=Fun, and Filled with awesomeness.
Exp= Debian Edu, you'll get better here. big_smile

Offline

#10 2013-04-09 05:05:13

VastOne
#! Ranger
From: #! Fringe Division
Registered: 2011-04-26
Posts: 10,163
Website

Re: When is testing more stable than stable (if ever)?

^ Thats about the best analogy of Debians ridiculous naming conventions as I have heard... Well done


VSIDO | SolusOS

Words That Build Or Destroy

Offline

#11 2013-04-09 06:25:18

zalew
#! Junkie
From: Warsaw, .PL
Registered: 2012-03-28
Posts: 374

Re: When is testing more stable than stable (if ever)?

DebianJoe wrote:

Debian's naming system causes a bias against the reality of volatility.

debian takes real effort into trolling the community with names. rejecting firefox name because of trademark => stupid questions about iceweasel vs firefox forever.

Offline

#12 2013-04-09 07:05:08

pvsage
Internal Affairs
From: North Carolina
Registered: 2009-10-18
Posts: 11,949

Re: When is testing more stable than stable (if ever)?

Perhaps it is time for another discussion on the meaning of the term "stable" as it is used in Debian?

Stable simply means "resistant to change"; a diamond is more stable than a lump of anthracite.  (Damn, I spelled that right on the first try - how nerdy am ? yikes )  Testing means "what makes the grade makes it into the next release"; if a particular package can't be made release-ready during the freeze shakedown (glares at minitube), it fails the "test" and is dropped.  Unstable is the opposite of stable; it is prone to frequent change...just like water in the classic Bruce Lee quote.


I'm a moderator here.  How are we doing?  Feedback is encouraged.

Offline

#13 2013-04-09 19:45:17

vicshrike
#! by Default
From: /grill
Registered: 2009-02-13
Posts: 2,960

Re: When is testing more stable than stable (if ever)?

this is the most recurring thread topic here @ the forums, it deserves a mega thread, if not a sticky smile


#!, all else is but a shadow

Offline

#14 2013-04-10 03:14:03

sunfizz98
Carbonated Orange Juice
From: su terminal
Registered: 2011-05-12
Posts: 1,876
Website

Re: When is testing more stable than stable (if ever)?

@vastone
Wouldn't running a server on sid just require more frequent updates? I'm sure it would be "safer" than the debian squeeze server, but not recommended for the forgetful.

Offline

#15 2013-04-10 05:35:11

zalew
#! Junkie
From: Warsaw, .PL
Registered: 2012-03-28
Posts: 374

Re: When is testing more stable than stable (if ever)?

sunfizz98 wrote:

@vastone
Wouldn't running a server on sid just require more frequent updates? I'm sure it would be "safer" than the debian squeeze server, but not recommended for the forgetful.

safety is not a matter of release schedule.

running a server on testing/unstable is not about being forgetful, but who and what you run this server for. a rolling release may only make sense if you are running it for yourself, because no customer will want to swallow the costs and consequences of your frequent upgrades. hence providers run marked releases of debian or ubuntu. even if you run this for yourself, rolling updates make your testing and deployment strategies more complicated and increase the points of failure. so, summing it up, for me it's a -1 for running a rolling release in production.

Last edited by zalew (2013-04-10 05:37:18)

Offline

#16 2013-04-10 12:51:05

VastOne
#! Ranger
From: #! Fringe Division
Registered: 2011-04-26
Posts: 10,163
Website

Re: When is testing more stable than stable (if ever)?

^ True... Although, it could be said that once a sid based server is setup and working correctly, it would never need to be updated (in theory)

Not meant to be an argumentative statement... I have just seen several production stable based environments left in a corner and not touched for 4 years ...


VSIDO | SolusOS

Words That Build Or Destroy

Offline

Board footer

Powered by FluxBB

Copyright © 2012 CrunchBang Linux.
Proudly powered by Debian. Hosted by Linode.
Debian is a registered trademark of Software in the Public Interest, Inc.

Debian Logo