I dont understand the compat-centric decisions being made for major version changes. It's as if the user-base count is the #1 priority. Many gridlock arguments because of this :(
Things as simple as fixing ternary associativity that has been broken and discouraged forever are basically DOA because it may break some tiny % of code that relies on broken behavior. They have a valid point that simply fixing it will break code in silent ways. So there was a possibility of making it unassociative in 7.0, throwing a warning and only fixing it for real in 5 years during the next major cycle. I don't know if this migration path will even happen.
>I dont understand the compat-centric decisions being made for major version changes
Compatibility is the only thing PHP has going for it right now. It feels like the only PHP apps out there are 'legacy' ones, new projects are written in newer stuff.
cweagans is right. PHP is deeply unfashionable, ridiculous even to some programmers using other langauges. But quietly, somewhere out of sight, it has been enjoying something of a renaissance in recent years. Generally I'm pleased with changes in the culture, tools, standards and even in the language. But just can't get behind this decision to keep mcrypt.
Seems like he just enjoys getting mad at people about their decisions. There's reasons for this decision, and as a PHP developer, I see why tearing out mcrypt could be problematic. Pretty lame that he's shitting all over these guys because they did their job and made a judgement call.
It appears those with a vote regard adoption of PHP 7 as the higher priority and therefore compat with existing code using mcrypt must be ensured. http://news.php.net/php.internals/82191
I haven't used php, but the situation described at that link is codependence.
a: "Most of our users don't care about security."
b: "OK then they can continue using old broken versions."
a: "No, then they won't be secure! Therefore it must be easy to upgrade."
b: "How?"
a: "By not making the proposed security improvements."
This seems like a recipe for losing any users that do care about security, which is not a viable strategy over the long term.
Things as simple as fixing ternary associativity that has been broken and discouraged forever are basically DOA because it may break some tiny % of code that relies on broken behavior. They have a valid point that simply fixing it will break code in silent ways. So there was a possibility of making it unassociative in 7.0, throwing a warning and only fixing it for real in 5 years during the next major cycle. I don't know if this migration path will even happen.