I've been happy with some of the great improvements to Ruby in the past year. But this one really leaves me scratching my head. It seems like a lot of complexity for what I perceive as a negative value feature. I'm concerned about damage to the gem ecosystem by encouraging gems to bundle version locked dependencies. The last thing I want is a gem deciding to have multiple versions of other gems in my app.
One of the tradeoffs that (imo) has been net positive in the Ruby ecosystem, is how a project has to ultimately load and run a single set of versions of all dependencies. It creates some extra maintenance work on one hand, but the result is that it encourages the ecosystem to not have to face the hell of having 3 or 5 or 10 versions of each common dependency within a project.
I recognize that this is an occasional cost to library maintainers, but in the long-term has contributed to benefits. My perception is that Ruby libraries have smaller list of dependencies than similar libraries in some other languages. There are several reasons for this, but this ecosystem pressure to stay compatible with a range of dependency versions is one of them.
It feels to me like this leads to a sweet spot for small- and medium-sized projects, and I can see it might have an upper limit? I have not been in the situation, but heard of situations where the largest projects inevitably run into needing specific versions of two separate libraries that don't agree on a version of a common dependency.
