You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now we don't distinguish between whether a registry was added explicitly by the user via registry add X or it was automatically/implicitly added because the user didn't have any registries added and the package server offers it. This leads to various confusions:
If the user happens to add a registry first, then they won't get any registries from the package server.
If what registries the package server offers changes, then the old ones linger and the new ones may not be picked up.
Instead, I would propose that we have two kinds of registries:
Explicitly added registries, which are always considered until the user explicitly removes them.
Implicitly added registries, which are only considered if the current package server offers them.
We could consider gc'ing implicitly added registries if they haven't been offered in a long time, so we may want to keep track of the last time we talked to a package server that offered an implicitly added registry. This would alleviate the above issues:
If a user adds a registry first, it doesn't matter—they'll use that as well as whatever registries their package server offers and the order of events doesn't matter.
If the user switches package servers or what registries a package server offers changes, we'll only consider packages included in the explicit registries as well as the ones included in the registries offered by the current package server.
Thoughts on this approach. Spurred by having long wanted this and conversation with @KristofferC and @Keno.
The text was updated successfully, but these errors were encountered:
Right now we don't distinguish between whether a registry was added explicitly by the user via
registry add X
or it was automatically/implicitly added because the user didn't have any registries added and the package server offers it. This leads to various confusions:Instead, I would propose that we have two kinds of registries:
We could consider gc'ing implicitly added registries if they haven't been offered in a long time, so we may want to keep track of the last time we talked to a package server that offered an implicitly added registry. This would alleviate the above issues:
Thoughts on this approach. Spurred by having long wanted this and conversation with @KristofferC and @Keno.
The text was updated successfully, but these errors were encountered: