Porting the files.sed operation?

[rpkilby]

Is there any plan on porting the sed operation from Fabric v1?

[bitprophet]

Not sure I made the call one way or the other yet. In general, at this point in time it's generally considered a lot safer and more reproducible to upload (potentially rendered) files wholesale instead of trying to manipulate them on the fly, which is certainly one reason why I personally haven't needed it.

OTOH there are always funky sysadmin reasons to occasionally perform manipulations instead of overwrites.

OTOOH I remember sed having an enormous number of irritating bugs for such a comparatively small amount of code, so I'm not in a huge rush to re-invite that into the fold.

OTOOOH many of those bugs were tied to how we performed text escaping because run et al used a shell wrapper, which is currently not the case in Fab 2.

It's complicated 😂 but I'll leave this open as a "maybe" for now, esp if somebody drops a PR and has pre-emptively written a ton of tests drawing from the old sed's bug list...

[rpkilby]

That all makes sense. The case where this seems relevant to me is modifying a few pre-existing system configuration files. e.g., ensure that sshd disallows root login and password authentication while leaving the rest of the conf at its default values. That said, maybe it would be better for us to track the entire sshd_config. Then again, mucking with the ssh conf isn't my responsibility anyway so ¯\_(ツ)_/¯