@jmoiron @TheMerovius Acknowledged, I just don't think consistently breaking software is a price worth paying that incentive. I like the idea that updating components is deliberate, because I think people underestimate the cost of doing so responsibly and we just get by on best intention...
jmoiron plays the blues
@jmoiron the best way to upgrade a dependency remains to remove it
@jmoiron a big problem I've always had with semver as its used today is that its proponents seemed preoccupied with how to make upgrading libraries easier and ignored the issue of how you might keep software stable while doing so
@jmoiron minimum version selection has the potential to be Go's 2nd biggest gift to the wider programming community (after go fmt); it was worth the wait: https://t.co/wiiUZeafOv
@jmoiron @dgouldin TOML might be the closest; still no schema, no formatter, some weird overall document structure ([[foo]] and [[https://t.co/Ts4WGgiqOC]]).. there's also still no official language freeze or 1.0 version though it's close. It nails most of the rest including not being ya...
@jmoiron config format reqs: no parser ambiguities, obvious structure, clear base datatypes (at least json types + integers), comments, multi-line strings, formatter, TOOWTDI, built-in optional schema (eg num ranges, enums), no dumb grammar rules that blow up the whole parse, not yaml
@jmoiron @bryanl @copyconstruct @bryanl your tweet about runbooks being a list of things to fix still resonates; a runbook can be an engineering tradeoff of things that are too expensive to automate at the moment, but if things stagnate it gets institutionalized as "the way to repeat pro...
@jmoiron @davecheney @dgryski and yet when we dig deep we see modern computing kept afloat by a series of caches, delicately balanced https://t.co/uBwBRIFKcS
@jmoiron @clofresh ambitious post id keyspace
@jmoiron @breerly @davecheney I don't think "under no circumstances are tracebacks acceptable", methods to prevent them often end up over-eating errors. Tracebacks are just not a replacement for error handling. This is poor: https://t.co/rtyy6XfMvZ
@jmoiron Tracebacks are not error messages and they do not belong in error messages. Do not show tracebacks to your users, even if your users are programmers.
@jmoiron every unhappy on-call rotation is unhappy in its own way
@jmoiron @copyconstruct Survivor bias. Companies who obsess over automation sacrifice things more vital to their revenue driving areas.
@jmoiron @francesc @sourcedtech reminds me of an airbnb I stayed at in paris with a ... peculiar bathroom https://t.co/VVg0UO03UC
@jmoiron the nice thing about eventual consistency is that it can be implemented on top of any datastore
@jmoiron @chimeracoder I'm pretty sure at some point in the past I printed a document and then scanned it so I could get it in PDF
@jmoiron @copyconstruct The etiquette is: if you think you can make it, feel free to cross + caveat emptor. If you can definitely make it, you must cross. I think the grid encourages this as it makes for a lot of intersections.
@jmoiron @copyconstruct @Medium yeah, sounds great!
@jmoiron @copyconstruct @Medium 2nd paragraph describes my experience with twitter, ironically
@jmoiron @RayS ditto
@jmoiron Imagine a world where it is 2017 but this type of replication didn't already exist. Who would be arguing for it to be called master/slave? Probably just assholes. https://t.co/go9a29RANI
@jmoiron Think about code coverage tools like you would infrastructure observability tools. They provide very specific information, an incomplete and sometimes misleading version of reality, just like every other observability tool, but even a little insight is way better than none.
@jmoiron new years resolution: fail any change request in code review that features a new type starting with the word "Abstract"
jmoiron/dotfiles 809b373 i like materialtheme and i dont like material so this messes up tab completion for no benefit
@jmoiron @dgryski twitter's weird concept of a "timeline" made it seem like a recruiter sent you that tired rob pike quote :D