There are changes that require a new edition, particularly around new keywords.
However, to be clear, editions are primarily a marketing/communication/project management tool. They give us a way to try to bring together a number of threads of work into a coherent whole, with a high level of polish across the board (including docs and tooling), and then to present that work to the world with a clear story.
It's worth comparing this to the recent Firefox Quantum release, which was likewise a singled out release on the normal train process that brought together a number of important changes, marketing/communication, and a high level of polish.
That's what I've been afraid of. When epochs were announced, I thought it is kinda nice to have backwards compatibility while introducing breaking changes. It's a good technical concept. Trying to change it's role into "marketing/communication/project management tool" out of a sudden is what puzzles me.
> It's worth comparing this to the recent Firefox Quantum release
I don't think it's a fare comparison. Firefox Quantum doesn't involve much of breaking changes (if you don't consider deprecating the old plugins one, that is). And the changes are most drastic since early versions of FF. With Rust though, it's clearly improving every day, I don't see Q3 2018 as any sort of "quantum leap". At least for as long as it doesn't include the const generics :)
> Firefox Quantum doesn't involve much of breaking changes (if you don't consider deprecating the old plugins one, that is).
They weren't just deprecated; they stopped working.
> I don't see Q3 2018 as any sort of "quantum leap"
For people following Rust closely, the Rust 2018 release won't be so special. But most people aren't keeping up with the details of new releases every six weeks; for them, we have an opportunity to take stock of what's changed over the last couple of years, explain the impact on idioms and how to transition code.
Remember that Rust is fairly unique in having a six week release cycle. Most other languages have a much slower release cycle, where every release is "major". Editions give us a way to pull together the rapid work we're doing into a coherent story.
From a management perspective, it's also a very useful way of focusing our effort as a community, to make sure all the pieces we've been working on are coming together into a polished whole on a clear timeline.
> That's what I've been afraid of. When epochs were announced, I thought it is kinda nice to have backwards compatibility while introducing breaking changes.
That is what editions are, though (though as I keep insufferably emphasizing, all breakage is opt-in). Can you be more specific about what you're afraid of?
> Firefox Quantum doesn't involve much of breaking changes (if you don't consider deprecating the old plugins one, that is).
You must not have been on any web forum discussing Firefox 57 if you don't remember the cacophonous, world-ending drama that deprecating legacy add-ons created. :P
However, to be clear, editions are primarily a marketing/communication/project management tool. They give us a way to try to bring together a number of threads of work into a coherent whole, with a high level of polish across the board (including docs and tooling), and then to present that work to the world with a clear story.
It's worth comparing this to the recent Firefox Quantum release, which was likewise a singled out release on the normal train process that brought together a number of important changes, marketing/communication, and a high level of polish.