[TYPO3-dev] New version scheme for system extensions

[Jigal van Hemert]

Hi,

On 16-12-2011 18:09, Ernesto Baschny [cron IT] wrote:
> in my (personal) opinion I prefer to stick to version numbers which
> makes sense for each individual extension. I see no reason to just "keep
> rising" version number of extensions on every patch level release even
> if nothing changed. Even between minor updates (4.5 -> 4.6), if there
> has been no changes in a particular sysext, why upgrade their version?

One very simple reason: it is much easier to combine the version of the 
sysext with the version of the core it was shipped with. Nothing more, 
nothing less.

A few arguments:

- people ask questions in a list posting and mention that it happens in 
Extbase 1.x.y or felogin 1.q.r; it is quite hard to see in which 
version(s) of TYPO3 this was included

- because sysexts are shipped with the core there is no need to have 
different versions than the core they belong with. It is neither 
encouraged or supported to use sysexts with different versions of the 
core than they were shipped with. Why should they have a different version?

- if someone has installed a local version of a sysext it will be easier 
to spot the version difference (htmlarea 4.5.8 on TYPO3 4.6.2 is easier 
to see than 1.2.3 which should be 1.3.1)

- sysexts actually should be updated for each release to reflect the 
TYPO3 core version they depend on; if this is automated the sysext 
version can also be updated.

> I have seen site integrators comparing sysext version numbers after an
> TYPO3 upgrade just to see if something "changed" (i.e. felogin
> extension). So they have also a good use as "stand alone" numbers.

We already supply a changelog, complete CVS history, bugtracker with 
categories for each sysext, etc. Honestly, I have never heard of anyone 
taking notes of sysext versions before and after and upgrade to see if 
sysexts were changed.

> So +1 to do it for Extbase (and Fluid as a consequence) but -1 for doing
> it on *every* sysext (or even to "automate" this version increasing on
> release time), because it gives site integrators false impression of
> "major changes".

 From this I must conclude that someone actually checks if there is 
anything changed in for example felogin between two (minor) releases and 
increases the version number?
That would be too much effort for very little (if any) reason.

I still feel that we should make it simpler, both technically 
(automating the version numbers) and for users (sysext version equal to 
core version).

-- 
Kind regards / met vriendelijke groet,

Jigal van Hemert.