Closed tgraham closed 12 years ago
I guess we'll have to agree to disagree because what might be considered sane for one person might not work for another so great leaps are being taken when assuming a default is sane.
I wouldn't want to classify all users as wanting unicorn and expecting them to deploy to heroku, by default.
Sane is what works in 90% of cases. Surely, webrick as a default production server is weird, right? And terrible for people who aren't rubyists and want to be able to deploy without having to edit anything.
I would agree with thin as a good alternative to Unicorn that should be an effective default. Webrick is simply unusable as a production server.
I pretty much merged this as part of merging #665. Thank you @tgraham!!! :star: :heart:
Bikeshed comment for puma here ;)
Given https://blog.heroku.com/archives/2013/2/16/routing_performance_update/ thin is actually significantly worse.
Worse than Unicorn? Well, yes, with Rails the way it is now that's true. That's why it originally had the comment "Strongly encouraged" :) But, it was part of our needs to run on Windows story, and this is the world we live in. Yanno? Our defaults were encouraged to be 'easy' and 'ubiquitous' with an assumption that nodes will not require severe concurrency and thus Unicorn is not as acceptable of a default. An extra commit was recommended here to switch to Unicorn when those assumptions are wrong, for instance, for a large-scale node such as the main rstat.us instance.
I submitted a PR specifically to discuss the issue further. ;)
Fixes Issue #547