> Sorta along the lines of the mantra "Don't design your code for extendability, design it for replaceability" (not sure where I read that).
I agree in principle and strive to do that myself, but it has almost never been my experience with code written by others across bunches of projects.
Anything developed in house without the explicit goal of being reusable across numerous other projects (e.g. having a framework team within the org) always ends up tightly coupled to the codebase to a degree where throwing it away is basically impossible. E.g. other people typically build bits of frameworks that infect the whole project, rather than decoupled libraries that can be swapped out.
> The upside of "what a few people within the org could achieve" is that a couple of devs spending a few weeks on a project are never going to make something that cannot also be replaced by a different couple of developers of a similar timeframe.
Because of the above, this also becomes really difficult - you end up with underdocumented and overly specific codebases vs community efforts that are basically forced to think about onboarding and being adaptable enough for all of the common use cases.
Instead, these codebases will often turn to shit, due to not enough people caring and not being exposed to enough eyes to make up for whatever shortcomings a small group of individuals might have on a technical level. This is especially common in 5-10 year old codebases that have been developed by multiple smaller orgs along the way (one at a time, then inherited by someone else).
Maybe it’s my fault for not working with the mythical staff engineers that’d get everything right, but so don’t most people - they work with colleagues that are mostly concerned with shipping whatever works, not how things will be 5 years down the line and I don’t blame them.
I agree in principle and strive to do that myself, but it has almost never been my experience with code written by others across bunches of projects.
Anything developed in house without the explicit goal of being reusable across numerous other projects (e.g. having a framework team within the org) always ends up tightly coupled to the codebase to a degree where throwing it away is basically impossible. E.g. other people typically build bits of frameworks that infect the whole project, rather than decoupled libraries that can be swapped out.
> The upside of "what a few people within the org could achieve" is that a couple of devs spending a few weeks on a project are never going to make something that cannot also be replaced by a different couple of developers of a similar timeframe.
Because of the above, this also becomes really difficult - you end up with underdocumented and overly specific codebases vs community efforts that are basically forced to think about onboarding and being adaptable enough for all of the common use cases.
Instead, these codebases will often turn to shit, due to not enough people caring and not being exposed to enough eyes to make up for whatever shortcomings a small group of individuals might have on a technical level. This is especially common in 5-10 year old codebases that have been developed by multiple smaller orgs along the way (one at a time, then inherited by someone else).
Maybe it’s my fault for not working with the mythical staff engineers that’d get everything right, but so don’t most people - they work with colleagues that are mostly concerned with shipping whatever works, not how things will be 5 years down the line and I don’t blame them.