You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For ES Module users, we could, for free, expose all common feathers server dependencies as ES Module exports under package.json. CommonJs users would be unaffected, but module users could drop their dependency count and improve DX by not having to run install as often if they want oauth, sockets or schema for instance. It also makes managing our package.json easier, and updating feathers is a single line affair, rather then a 10 line commit every time.
Again, this is free real state, it would not have any runtime overhead. 200K extra of hard drive space is negligible in a server context, even in the most frugal environments Feathers runs on.
The only downside is it soft-locks us into supporting a module for future versions... but I don't see socketio, oauth, adapter-commons or any of the top 8 modules non-express modules going away in the relevant future.
I could make a PR and ReplIt sandbox to show this in action if this sounds good to everyone.
p.s. I think we should do to Koa/Socket.io what Laravel/Vite did to Symphony/Roll-up.
Related (this would have been another package): #2522
Having @feathersjs/server, @feathersjs/common, or @feathersjs/omnibus would be just as convenient while keeping reproduction sizes small and protecting the bundle size of exotic or ancient client side bundlers.
The text was updated successfully, but these errors were encountered:
For ES Module users, we could, for free, expose all common feathers server dependencies as ES Module exports under package.json. CommonJs users would be unaffected, but module users could drop their dependency count and improve DX by not having to run install as often if they want oauth, sockets or schema for instance. It also makes managing our package.json easier, and updating feathers is a single line affair, rather then a 10 line commit every time.
Again, this is free real state, it would not have any runtime overhead. 200K extra of hard drive space is negligible in a server context, even in the most frugal environments Feathers runs on.
The only downside is it soft-locks us into supporting a module for future versions... but I don't see socketio, oauth, adapter-commons or any of the top 8 modules non-express modules going away in the relevant future.
I could make a PR and ReplIt sandbox to show this in action if this sounds good to everyone.
p.s. I think we should do to Koa/Socket.io what Laravel/Vite did to Symphony/Roll-up.
Related (this would have been another package): #2522
Having @feathersjs/server, @feathersjs/common, or @feathersjs/omnibus would be just as convenient while keeping reproduction sizes small and protecting the bundle size of exotic or ancient client side bundlers.
The text was updated successfully, but these errors were encountered: