On Wednesday, 22 May 2013 at 13:44:10 UTC, Dicebot wrote:
Eh, official definition of "breaking change" keeps breaking my
heart. But I guess this is a mindset set in stone now and
changing it is close to impossible.
Can you elaborate a little bit?
I felt personally that what the discussion didn't really address
was that these days the "breaking" problems are primarily not
with the language, but with the standard library, and this is not
much less problematic. Yes, it was mentioned, but not really
considered in detail, and that's where a significant amount of
attention needs to be.
You can extend this to spec -- whereas D as a language is
substantially well defined, the standard library doesn't seem to
be, and there are clearly areas where this needs to be addressed.
Just as one example, which strongly affects my own use of and
engagement with D: the de facto behaviour of random number
generators, which are implemented as value types despite there
being strong reasons why they should instead be reference types.
I understand that resolving the last major language spec issues
probably needs to be first priority, but I don't think this is
the major _usability_ issue any more.
