[JohnBooty]

I'm a Rails developer and I'm coming to feel this way as well. I agree with most of what's in the Rails Manifesto except the convention-over-configuration thing. Too much "magic" happening. Magic's cool, but I need to know why a certain thing works or why it doesn't work.

Rails is the only environment in which I've routinely found myself utterly stumped. Not because it's hard, because it isn't. But the thick layer of "magic" makes it awfully hard to debug sometimes.

[jmagoon]

We are in the same boat. I find myself relying way more on pure ruby classes.

It's also dangerous because Rails fits a specific niche, like he mentions--building (potentially) many web systems quickly. We chose to migrate large legacy systems to rails, which meant lots of the nice convention just. doesn't. work. Day 1, you're already digging in to figure out how to get around the magic!

[ansonhoyt]

For me, magic isn't the problem. I like the magic of conventions, but not when it is painful to pull back the curtain. For Rails, I think pain comes when magic stands on poor internals, which gets amplified when gems layer on their own painful workarounds...and I have to sort through all of it.

That's why I'm excited about Sean Griffin's work on the [Attributes API](http://edgeapi.rubyonrails.org/classes/ActiveRecord/Attribut...). To do it, he's had to refactor out a lot of internal monkey patching so that (in Rails 5) there is only one source of type information for an AR model. This means gems can extend AR, I can extend AR, new requirements can walk into my door that I can deliver on without monkey patching or opening Pandora's box of a patch of crappy Rails internals. The magic is still there, but it's done it a way so the magic doesn't fall apart in unexpected, unmanageable ways. That's what matters to me.