Bump Mail dependency to pick up downstream memory improvements

Hey all,

Without wanting to claim the hard work of others!! I would love to create an issue that tracks bumping Mail to ~> 2.6 for Rails 5 (e.g re-create a PR like this one). Here’s why:

Jeremyevan’s PR to use a text-based columnar storage in the Mime-types gem has been shown to significantly decrease memory when a user opts in to it (I’ve seen 20mb decrease personally). This memory saving behavior is a default in the 3.0 release of Mime-types (as well as a raft of other, really lovely changes). Given that Mail works with Mime-types 3.0 and when Mail 2.6.4 has finished brewing it’ll allow Mime-types 3.0 It would be great for Rails to nudge users to pick up this behavior too.

What are people’s views on this? It feels a little audacious of me to modify one line of ActionMailer’s gemspec but I’m really keen to see this change happen.


It is already allowed to use mail 2.6 with rails, so why should we disallow to use mail < 2.6?

That's a solid point. I don't really have a response other than because of the benefits above and the opportunity of doing this during a major release (e.g when people are already upgrading gems). Logically, in my head seeing a memory decrease during a major release of Rails by being progressive with a non breaking dependency makes it worth ~> 2.6

No harm either way as eventually, people will catch this update :).

That is a great point too. I can see both sides. But looking at the memory decrease as a “feature” seems like a winner to me to at least try to go ~> 2.6 for the initial release of Rails 5.

  • Ken

Thanks guys. I’ve opened up a PR. Let’s see what happens :slight_smile: