- Darcy Clarke (@darcyclarke)
- Isaac Z. Schlueter (@isaacs)
- Rebecca Turner (@iarna)
- Nathan LaFreniere (@nlf)
- Nathan Fritz (@fritzy)
- Jordan Harband (@ljharb)
- Gar (@wraithgar)
- Tierney Cyren (@bnb)
- Owen Buckley (@thescientist13)
- Luke Karrys (@lukekarrys)
- Rick Markins (@rxmarbles)
- Housekeeping
- Introduction(s)
- Code of Conduct Acknowledgement
- Outline Intentions & Desired Outcomes
- Announcements
- Issue: #472 [RRFC] Only use registry provided URLs to fetch tarballs - @iarna
- Issue: #473 [RRFC] npm outdated (npm >= 7.24.2) exit with incorrect status code - @nassau-t
- Issue: #471 [RRFC] Exclude node_modules from Time Machine backups (macOS) - @probablykasper
Issue: #472 [RRFC] Only use registry provided URLs to fetch tarballs - @iarna
- @iarna
- scoped packages cannot be installed in certain situations/scenarios, specifically when the registry destination is replaced
- this is a change from v6 to v7
- scoped vs. unscoped modules behaiving differently is definitely a bug
- replacing/rewriting the registry destination is definitely a bug
- in this proposal
- the registry that is used to create the lockfile will get stored
- the lockfile should be recalculated if the registry changes & there's an integrity mimsatch/invalidation
- @isaacs
- agree
- the registry rewriting was incosistent prior to v7, we doubled down on standardizing on some magic swapping
- should track the registry in the lockfile
- need to ensure we nerf-dart the registry value before storing it back in the lockfile
- @ljharb
- would have been nice to get ahead of this & have a single entry for unscoped/bare packages & then another for each scope (reducing lockfile churn)
- @isaacs
- an interesting scenario we should consider is when someone installs
npm install foo --registry=bar
& then does the entire tree need to be resolved again based on that new registry? - suggestion: could consider this scenario to be a good use-case for scopes (ie. if you have a subset of dependencies that resolve from a different registry then use a scope)
- an interesting scenario we should consider is when someone installs
- @bnb
- what is the structure being proposed/changed?
- @isaacs
- any changes to
packages
/dependencies
need
- any changes to
- @bnb
- how doe scoped registries apply in this?
- @isaacs
- the proposal is currently just a single registry
- @bnb
- perhaps we could have a duplicate of the single registry to dictate how scoped registries resolve
- @iarna
- that could work
- @isaacs
- using something like
@foo:registry:<url>
would work. (paraphrased code example)
- using something like
- Actions:
- @iarna to make RFC/PR
- @darcyclarke to backlog work for lockfile changes
Issue: #473 [RRFC] npm outdated (npm >= 7.24.2) exit with incorrect status code - @nassau-t
- @wraithgar
- we should have this decision documented
- @isaacs
- old versions of npm (ex. v6) exited 0 & we broke this in v7 & then fixed it (ie. this is working as intended)
- @ljharb
- we should document npm's ideologies
- @iarna
- wierd/nuance which deviates from some of these standards/ideologies
npm
should have is that: the error code does not get set if you pass--parseable
or--json
- wierd/nuance which deviates from some of these standards/ideologies
- Actions:
- @darcyclarke create a doc that outlines npm's ideologies
- @darcyclarke close RFC & other issues related to exit code statuses, referencing this new document
Issue: #471 [RRFC] Exclude node_modules from Time Machine backups (macOS) - @probablykasper
- @ljharb
- this has been a pain for awhile
- @darcyclarke
- should this be opt-in?
- even though we think no one relies on this today, there always is someone that ends-up being broken by a change like this
- @bnb
- personally prefer opt-out
- Actions:
- @darcyclarke backlog work item to investigate this work & initially ship under a minor in v8 as opt-in & v9 as opt-out