LLVM 4.0 Planned For Release At End Of February, Will Move To New Versioning Scheme

[phoronix]

Phoronix: LLVM 4.0 Planned For Release At End Of February, Will Move To New Versioning Scheme

Hans Wennborg has laid out plans to release the LLVM 4.0 (and Clang 4.0, along with other LLVM sub-projects) toward the end of February...

http://www.phoronix.com/scan.php?page=news_item&px=LLVM-4.0-Release-Schedule

[gufide]

Version numbers are becoming meaningless... why the change??

[Mystro256]

Honestly it makes more sense that way. 3.9->4.0 wasn't any different from any other minor version bump. They were literally just adding 0.1 every release every time; the major version number bump literally meant nothing.

They're the latest project joining the big version number race

Calling a race is just silliness. I would think most projects that are switching this scheme aren't doing it for histrionics, but rather because they realized the major version was pointless and see no reason to keep both a major and minor number. Bumping the major instead of minor then the new minor becomes the patch version.

Side note, I still don't understand why chrome/chromium is MAJOR.MINOR.BUILD.PATCH.... Do they even bump the minor version? I've always seen it as 0, e.g. 55.0.2883.75.

[dungeon]

Version numbers are becoming meaningless... why the change??

Marketing reasons are only i could guess With that, development virtually looks like more rapid than it really is

[Marc Driftmeyer]

Having read that thread I don't see a new versioning scheme. I see push back from devs asking who the idiot was that came up with it.

[waxhead]

Version numbers are becoming meaningless... why the change??

Could not agree more.

What is wrong with VERSION . REVSION . BUILD . PATCH . LETTER

VERSION:
This is a fundamental change. E.g. a program rewrite from scratch or other large changes like a new GUI etc...
may or may not be compatible with the previous version.

REVISION:
This is an adjustment to the version , introducing new features and should remain compatible.

BUILD:
This is for bugfixes in the revision

PATCH:
This is for fixing critical security bugs etc...

LETTER:
usually something along the lines of....

A = Alpha
B = Beta
RC-X = Release candidate-X

I agree with those who say that big version numbers is marketing only. What about firefox? it probably reached version 50-60 or something by now. What meaning does it have anymore?! It says nothing about anything at all and it would be far better to use year.month as a versioning scheme for fast moving targets even if I really don't like that either. For exampel a program 10.02 may appear to be outdated, but if there is nothing wrong with the program what do you need to update it for?! Version 10.02 may for all you know be more stable than version 16.12 and version 16.12 may not really have any new features you need either... the point is that you have no idea / feel about what the version number indicates!!!! Version 1.2.12 is nice since it says that there have been 12 fixes for version 1.2 - ergo probably the feature introduced after V1.1 now really works well!

Why do we have to reinvent what works all the time!

[zeehio]

I bet once they reach a high enough number in the hundreds and nobody pays attention to it anymore, they will say that version numbers are useless and they will switch to a date based version number (year.month.patch)

I don't get this crazy major version bumping...

[M@yeulC]

Well, as long as it keeps increasing...

I am not sure to have noticed a consensus on the new versioning scheme on this thread, though.

[FishPls]

I must be the only one who really likes big version numbers as releases. The browsers started to do this as a means of advancing the web platform so there wouldn't be another IE6 situation again, and it worked out very well. Frequent releases that bump the version number make users want to upgrade more. The addition of auto-updaters was another major factor.

[ldo17]

I always felt that changes in smaller numbers seemed more significant than changes in larger ones. Thus, going from version 2.0 to version 3.0 seemed a bigger jump than from 9.0 to 10.0.

To try to maintain the same sense of drama, I thought it might make sense to follow a Fibonacci series for version numbers. Thus, going 2, 3, 5, 8, 13, 21 ... should maintain similar levels of expected significance.