On Saturday, 10 June 2017 at 23:30:18 UTC, Liam McGillivray wrote:
I realize that there are people who want to continue using D as it is, but those people may continue to use D2. Putting the breaking changes in a separate branch ensures that DM won't lose current clients as they can just continue using D2 as they are. Even after D3 is stabilized, D2 could have continued support until nobody wants it.
Here is my suggestion for how to migrate into D3 well, if it were to be done:
First define a "Small D2". The goal of this language is to not really be that much smaller(it could still leave in most of the language, in fact), but to sweep up all the changes breaking with D2 into a form that's clearly less flexible while still being palatable for the majority of modules. It's allowed to be intentionally clumsy and regressive in some ways so that it can be forward-looking later, but not so much so that people don't want to write in it.
Small D2 is given the freedom to do things like regress standard library functions back into hardcoded compiler behaviors, to be more restrictive, to allow less ambiguity, and to generally follow with existing idioms but break the things that needed breaking and leave out the parts that didn't work. The new stuff of Small D2 would be "one-trick-pony" features that bottleneck what can be expressed into a "library" that is actually a compiler hack, or "new syntax" that is actually fake and only allows arbitrary expressions that are already possible in other ways.
Most importantly, Small D2 retains compatibility with regular ("big") D2 modules, so codebases can be migrated into Small D2 piecemeal, while leaving a small remainder to be migrated in a larger way when D3 rolls out. In essence, Small D2 allows the underlying formulation of D to be challenged, but it actually relies on regular D2 constructs internally, so iteration on design can go faster than if we "did it for real". And because Small D2 gives people a taste of new stuff, it will have some instant appeal: People should want to write new code in it, even knowing that it's a hack and less powerful.
Small D2 would be the debate platform for what the "core" of D is. Then, D3 is Small D2 with everything "done for real" this time. Small D2 code migrates forward seamlessly and can continue operating as if nothing happened. The rest of it is considered breaking.
