Re:Haskell’s State monad is not a monad (2014)
by lmm on Tuesday July 03, 2018 @10:58AM
(#17450722)
> You sure do, but you pay someone else to deal with all that nonsense.
I don't think I follow? I've got a class of problems where I might want to restart my application, and a class of problems where I might want to restart the VM/machine. I don't see how adding containers in between the two helps.
> Any language-specific solutions like .war files have no answer for binary dependencies or utilities provided by a package manager, and they're entirely right not to try and solve that problem.
JVM culture steers away from depending on binaries on the host system; in the rare cases where you want JNI then something like https://github.com/mrburrito/jni-loader is the standard approach IME. I think I've used literally two system libraries in my career, and both were standard enough that I was content to include them in the OS baseline. Appreciate that other languages do approach this differently.
> The other advantage is your deployment artifact has a single identifier, so it gives you a much nicer story for rolling back a bad deployment.
I don't see how this is different from any other kind of deployment?
I don't think I follow? I've got a class of problems where I might want to restart my application, and a class of problems where I might want to restart the VM/machine. I don't see how adding containers in between the two helps.
> Any language-specific solutions like .war files have no answer for binary dependencies or utilities provided by a package manager, and they're entirely right not to try and solve that problem.
JVM culture steers away from depending on binaries on the host system; in the rare cases where you want JNI then something like https://github.com/mrburrito/jni-loader is the standard approach IME. I think I've used literally two system libraries in my career, and both were standard enough that I was content to include them in the OS baseline. Appreciate that other languages do approach this differently.
> The other advantage is your deployment artifact has a single identifier, so it gives you a much nicer story for rolling back a bad deployment.
I don't see how this is different from any other kind of deployment?