[TYPO3-dev] Extensions need update for 4.3!

[Jigal van Hemert]

Ernesto Baschny [cron IT] wrote:
> I am against to just "change it because we can" (to use the expression
> that has gained popularity these days...). The change that breaks
> backwards compatibility has to make some sense in a productive way.

Very well put!

It's now a discussion between two approaches. A programmer who wants to 
limit the number of problems by exactly defining what is expected and by 
enforcing those rules, versus an owner of a web site who wants to keep 
his site running and who understands that regular updates are necessary 
to minimize the chances on getting any problems.

The main problem with the core programmer's behaviour is that not the 
developer who built a badly coded extension faces the consequences of 
the core programmer's strictly enforced rules, but the owner of the 
website is the real "victim" here.
His site stops working after upgrading to 4.3.

"He should contact the author of the extension and have him fix his 
badly coded extension". Sometimes that works, but many extensions in the 
TER are not maintained on a regular basis.

"Then he should fix it himself and take over the extension". Many 
website owner cannot do this. They could hire a programmer to maintain 
the extension, but this would be very expensive.

Strictly enforcing some rules with such severe consequences can only 
harm the product TYPO3 and will have very little advantages. If you 
think a bit further, TYPO3 has a huge repository with way over 3000 
extensions. These are all presented on the TYPO3 website and they are 
available from within TYPO3 itself. This looks like they are approved by 
TYPO3.
With the publication of 4.3 it should announce: "this is the newly 
improved version of TYPO3 with 100+ new features and bugfixes. And by 
the way, we also managed to render a few thousand of the extensions 
useless..."

> There are some changes that are of course not backwards compatible,
> which will break extensions and which we did in the past (upgrade to
> PHP5, change of paths and symlinks), but in most case there was a good
> reason to do so, breaking the one or other existing extensions. In my
> eyes this is not a good reason.

Indeed, this action reminds me a bit of the sudden change in the 
C-library which is used for PHP functions like mktime(). In the 
documentation was the subtile statement that the Unix timestamp for 
dates before January 1st, 1970 were undefined. For years the function 
created negative timestamps, which were widely used. All of a sudden a 
developer decided to follow the documentation a bit stricter and the 
next version of that library did indeed produce an "undefined" result...
A very expensive change it was...


-- 
Jigal van Hemert.