Open Ballot: Do version numbers matter?

TuxRadar

We'll soon be recording podcast episode 12, and our big debate is about version numbers. After the fun and flamewars surrounding KDE 4.0 and KOffice 2.0 - major version number bumps for "developer-focused" releases - we're wondering if the system needs to change. Why is Window Maker still at version 0.92, despite being stable for years? Should we have more 1.0s to make free software appear more complete and mature? Or should we just scrap major/minor numbers and follow the lead of less, which is at version 429?

You should follow us on Identi.ca or Twitter


Your comments

Well, they are functional

The advantage of release numbers is that they are always easy to understand when referring to software development. They are not, however, that easy for the general public, which is why what Ubuntu, amongst others, does is useful, as not only do they have release numbers e.g. 9.x but also names for the whole version e.g. jaunty. However daft the name is it is easier than saying 9.153215654 (hyperbole). But the system of just naming it of that year is also useful, but can seem dated or limiting if more than one version is released per year. This missing out of release 13 (Micro$oft Word) is just inane though!

Sure, when they make sense

They don't matter when the project clearly outlines their roadmap for each.

With one exception being x.0 ones. I'd consider that always to be ready.

Yes, though AI software

Yes, though AI software should work from a high number to 1

Who cares?

Yes, version numbers document which features change (2.6.29, 2.6.30) but too many points make it pointless (0.1.1.3, ??)

Sometimes it is necessary (GPL 3) but GPL 3.0 doesn't help. See Linux Outlaws episode "Pointless use of Noughts". 1.0 is no different from 1, but the difference between 3.2 and 3.11 confuses some, as 3.11 is newer. (eg Glibc 2.10, 2.8, 2.10 is newer). So just as long as it doesn't get too confusing (two point ten, not two point one oh), it might help.

And as for rolling releases, I'm in favour. I've decided not to number my software, just do a changelog with the date.

Depends as always;p

First of all on the project and the way it's developed.

I'd personaly like something along the lines of Source.Binary.Revision, so if you have 1.3.5 and 1.3.6 you know they are binary compatible but 1.4.1 would require a recompile and 2.4.1 might require changes in your source code. However this doesn't make sense for anything other than libs or plugin interfaces (if it makes sense at all).
Also it has the same problem most x.x.x types have that at some point you'll probably just be increasing a single number into infinity.

For less it probably makes sense to have a single number since there won't be any new amazing features or complete rewrites. Whereas for KDE it makes sense to mark the difference between 3.x and 4.x.

All in all what probably needs to change is the attitude of people to bitch about stuff they shouldn't really care about or they simply don't understand (you don't understand unless you're part of the project). It doesn't matter what the number is. Just because it's 1.0 doesn't mean it's stable etc. The best thing that can be done is to listen to what the developers say, and not interfere with useless flamewars about the version and/or other stuff.

go back to the old ways

I remember when the minor number being odd told you that the software was stable; odd let you know that it was a development release.

I wish the KDE team working on 4 technically had a stable release with 4.0; it just was not feature complete. They fumbled upon communicating this.

With 4.1, everyone knew it was unstable and feature incomplete. With 4.2, I have found plasmoids and other basic features to be unstable. They should not have gone to a 4.2 release until it was tried and true.

Personally, it has gotten to the point with KDE that I don't care because I can do everything I need/want in Gnome, XFCE, or LXDE. I don't need KDE. I like some of the things they are doing, but my mind has categorized them as a "watch and see" kinda software.

Numbers

On the one hand, I'm inclined to think it doesn't matter too much, one just looks for a higher number to see if there's an update. On the other hand, I think a more consistent kind naming might be desirable.
On the whole, I'd go for using the realease date (somewhat the way Ubuntu do for distros) but in the form foo-yyyymmdd, with the possibility of adding the letter D for development releases and an S for stable releases.

On the subject of KDE and KOffice, I don't see why people have been getting so worked up about it, KOffice isn't integral to KDE, you can run KDE without KOffice and you can run KOffice on a different desktop, if you want to. Some people must have too much time on their hands.

Absolutely... because of version expectation inflation

A few years ago my family had a boat called the Javelin 200. This boat was, naturally, a mockery on the lake because everyone knew that to have a boat worth any salt you needed a name at least into the thousands. So no matter how fast she went, the Javelin was always instantly relegated to second class in people's minds.

I think it works the same way with software, in that we see generic inflation in what people expect over time, though it does seem to be context specific.

Office apps, for example, are now considered unusual if they haven't broken a 2 level release. Who was this guy that was late to the spreadsheet party? What formulas were hanging him up?

Other, less proto-typical apps do fine at 1 and pre-1 levels, because low numbers don't strike people as unusual in the same way. Compiz is a good example.

So, absolutely the numbers matter... because it's all about expectations. People can compare what's the latest in any scheme, but they'll compare across schemes too even if it doesn't make technical sense.

stop with the dumb polls

stop with the dumb polls already, write something that means something.

Yes, and it always will

It is impossible to not have version numbers to keep track of what build you are on. Even more so with the way that some distros work, the build numbers let you know what parts work with what.

Evolution is a prime example of this. While version 2.24 works with the Scalix connect plugin, version 2.26 doesn't. Evolution doesn't keep backwards compatibility with plugins when evolving. Knowing that you need to remove v2.26 and install v2.24 to get your plugin to work is useful knowledge.

To the general public they just want stuff to work and don't care if the version is 123.123,34,2.234.23.42.1234 or v1.

New numbering scheme required.

Consistensy is the key. I suggest everyone should shift to the the defacto numbering scheme, 'Base Redmond'.

Let's count to version 7 with Microsoft:

1, 2, 3, 95, 98, 2000, ME, XP, 2003, Vista, 7.

Anything else is just silly.

stop with the dumb polls +1

I totally agree with this one. Some of these "open ballots" are slightly missing reality.

"should we thin the licensing herd?" - it's not like you can do anything about it. Someone needs something different they make another one, some are useless copies they will be forgotten. The reason there are so many is that they are probably needed (so they will stay no matter what you say) or were probably needed (and you cannot "unwrite").

"do we need a standard package format?" - missing some of the reality of what a distro is (full comment in the respective open ballot and podcast).

They matter within projects, not without

Version numbers play a critical role within an individual project, but since there are no clear guidelines for projects to follow when deciding version numbers, they don't matter much when comparing between projects. I like the major.minor.bugfix versioning system much more than the less version system (Is version 429 a big change from 428? Do I really need to upgrade from 427? Does less even get big changes?) but what counts as major and minor can be very subjective.

I think projects shouldn't stay at sub-1 version numbers forever. Once the project is considered generally robust and ready to be used by it's target audience, it should hit 1.0.

stop with the dumb polls +2

This site has really gone downhill. It started out really promising with lots useful tutorials and information taken from it's parent magazine but nothing of interest has been posted here in months. Sometimes nothing new appears here for over a week, even then it's probably just a release announcement or a link to some other site's tutorial.

All the polls are doing is provide filler for the podcasts, which quite frankly I couldn't be bothered wasting my bandwidth on.

I've now removed this site from my RSS feeds as there are much better sites for news and links to tutorials.

stop with the dumb polls +3.1416

Every time I check the batch of Linux RSS feeds I have in Google reader, it's always "Why you should use Linux instead of Windows" or some inane "pool" that doesn't mean a thing or "10 apps to to replace your windows applications". Rince. Repeat.

I like Linux, I like the principle behind it, but isn't time you guys stop telling the world how great Linux is and actually start making it better?

Personally, I think this "There's a thousand Linux distro, but that's ok! It's about choice!" mentality is what is gonna kill the whole thing. It sure got me bored to death.

Linux needs(among other things)standards to grow and rid itself of it's "Linux is for nerds" reputation. Because, I'm sorry to tell you, but right now, Linux *is* for nerds. You guys like that? Good. Then stop whining about how the mighty Microsoft is stealing your thunder. People want something that works. That's it.That is all they want. Linux doesn't work. Linux is everywhere and nowhere at once. It's guts are spilled on the floor and no one seems to be able to figure out what to do with it. I can't even install a distro without fucking around a text file to get the right resolution. That should *not* happen.Never. Proprietary hardware/drivers you say? Well, them are the breaks. Deal with it. Fix it. Make it work *now*.

The year of the Linux desktop? Eh. Right. Never gonna happen until every one and their dog stop "building" their own distro (read: same goddamn distro with a different wallpaper) and actually start working *together* to build something relevant. I don't have a solution for this. I realize that's an ideal. Never gonna happen. Humans being what they are, we always think that we're better than the next guy and that we can make a better distro. It's counter-productive. But hey, that's the way it is.

Sorry to unwind like that. But it pisses me off to no end to see a great project like Linux being thrown to the four winds while nobody has a damn idea what to do with it or how to direct it toward a meaningful goal. I've had this on my mind for a while and I guess today is the it all comes out.

That's my 2 cents. Ok, maybe my 2$. :)

Peace to you all.

Stop with the dumb polls +4

Please.

But otherwise, keep up the great work!!

Re: Open Ballot: Do version numbers matter?

No, it doesn't matter.
It's just the developer's responsibility to come up with a suitable version number, not just randomising it up.

dumb poll ++

I think tuxradar is running out of good ideas. Hope you guys come back on track with interesting stuff instead of a poll for version numbering scheme ! honestly should we make a standard for versioning and call it Open Versioning Standard !!!?!!

Various comments

OK firstly I am theguywholikeslinux2 because the website has been very buggy recently and I can't log in.

The whole KDE flame war was basically over one version number. A. that should never happen. B. why couldn't developers just call it 3.9.x until it was stable (or should that be until it is stable?).

Giving releases names like Ubuntu does is a good idea. Or just call each release 1 number higher, end users aren't bothered whether it is .764.2467.01, revision 24576.343.23434. Only the developers need to know that!

2 numbers are ok, so 3.5 for firefox is fine (maybe it should be 4 but lets not go into that now) but why tell the users that it is 3.5.249.34 Beta 7.45? they don't want to know that!

Basically version numbers shouldn't matter but developers should make a better effort to name things!

Re: dumb poll, firstly this is not a poll! secondly it is a discussion!
OK, think about it, as I said earlier the whole KDE flame war was about a version number, if they named it correctly this open ballot wouldn't have happened. And as for the other polls (I mean open ballots) they have been quite reasonable. If you ask me I don't care how many distros there are but please give them the same package manager!! We need at least two distros to provide competition and innovation, but we can't spread the workload so thinly, we don't have enough time to re-invent the wheel! OK, tuxradar guys, listen to there comments, you may need a bit of improvement, but if you ask me you don't and you are the only feed in my feed reader (not including my Open University news one) and I don't intend to delete it any time soon. Keep up the good work and the podcasts, they are great!

Yes

Why should we change them? In my opinion they mostly work fine right now. Minor numbers *are* needed to show how much actually changed - I probably won't download v1.2.11 of a program - I'll wait 'til v1.3.

And anyway, how else would people be able to hit v3.14159265?

----
lucjon, formerly person-b
----
(Also, to all whiners on this thread, if you don't like Tuxradar, just stop visiting. OK, I'll stop feeding the trolls now.)

I think having a standard

I think having a standard version scheme is essential.
Major version numbers must be for major releases only. I must say that I hated the KDE 4 release plan. They should simply used 3.9 version until the product was completely ready. They broke the tradition in my point of view.

As I'm a Java developer (I know you may don't like it! ;) ) and specially a lover of JBoss products (a devision of Redhat) I can tell you that their version scheme is really good! For example consider JBoss Application Server. The major numbers are for major releases. The second number indicates some changes and we have a bugfix (I guess) number: e.g 4.2.3

The also include some RC, GA and ... at end to indicated Release Candidate, Generally Available, Beta and other form of release. I found it really useful.

I think some sort of standardization is always desirable unless we want to confuse the end user more.

This is my idea!

I think that all programs that follow an open-source license (i.e GPL or BSD, etc) should have a version system like KDE or Xfce. This is where there are three numbers. (i.e 4.6.1) The first number is the main version of the app, for example Gnome 2 or Kde 4. The second number is the version within that main version, for example kde 4.2 or Xfce 4.6. The last number is the minor version, i.e bug fixes and minor changes. For example Kde 4.2.1 or Xfce 4.6.1.

That way, everyone would know exactly what stage the programs at. For example, people would know that x.x.1 is more stable than x.x.0.

Use Rolling Release

If version numbers do matter to you, try to use rolling release like Gentoo or Arch Linux. Even Debian sid will do;-) Try to catch up with the latest stuff.

But who cares the versioning numbers, it doesn't mean the bigger the better;-)

annoying, but meh

It is silly to release anything other than an alpha/beta/updates less than a whole number release -like VLC 0.8.6. The first working version should be 1.0. After that, work your way up as features are added. Less than 1.0 really makes it look like just a silly college project with no wings.

KDE 4.0 was redonkulous in that even 4.2 has tonnes of small features missing - the buttons are there, but they are just grayed out or do nothing when clicked. -WHAT?!-

PS-I don't mind the silly polls. If it weren't for the silly poles, we wouldn't have polls to make fun of other's opinions!

PSS-quote:
"Let's count to version 7 with Microsoft:
1, 2, 3, 95, 98, 2000, ME, XP, 2003, Vista, 7."
HAHAHAHAHAHAHA!!!!!!!!!!!!! So True! -but I think 2007 is missing if you count 2003 -and can you even count ME?

Dumb polls - a suggestion

I agree this isn't the most inspiring open ballot yet to appear on Tux Radar, but the site is still excellent, and we should let the guys off just this once - it is holiday season after all.

I've thought for a while though that this site could use a general suggestions area, where readers can post any quesitons for further discussion. One such 'suggestion box' could be one where people can submit topics for the next open ballot.

Can this be done?

Yes, version numbers matter.

Yes, version numbers matter. But to appear friendly to new users, FOSS needs to cut the crap with 0.0.0.02.2342846alpha1 versioning system. VLC should have been at 1.0 a few years ago. When there is a "sub-1.0" software, end users assume it isn't ready. We need to change.

Ubuntu version is good for standar

I think ubuntu version number is good for standar how to use version number. So common user know that the software is stabil version and easy to understand

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Post new comment

CAPTCHA
We can't accept links (unless you obfuscate them). You also need to negotiate the following CAPTCHA...

Username:   Password:
Create Account | About TuxRadar