Nelly Yopa sent an Outreachy patch adding a promisor.fetchOrder config variable, a comma-separated list of remote names. Listed remotes are tried first in that order, then unlisted ones in config order, with the extensions.partialClone remote still last.
Christian Couder pointed to the Don't work on a proposed project right away section of the microproject guide. He said the patch addresses the main issue of a project the community is proposing, and that a design doc or long discussion should come before implementation. He gave an example of why: when fetching from different remotes, remote A may want the client to use some promisor remotes (A1, A2) while remote B advertises others (B1, B2, B3), and the best order may depend on which remote related objects came from, which a single list cannot express.
Kaartic Sivaraam asked Nelly to include the right people in To/Cc, and said it might be best to avoid premature work among Outreachy contributors before selections are out. He noted that if no candidate is picked for a project, working on it afterwards is fine.
Nelly said she had misread the guidance as applying only to projects someone else was already working on. She will leave the patch as an RFC prototype and wait for the selections. Junio C Hamano suggested that next year the project list might have three tiers: microprojects, mid-sized projects for demonstrating capability, and the real projects.
Two nearby threads touch the same topic. Hussain Sitamouwala, whose documentation patch on promisor fetch ordering was queued at v3 as hs/doc-promisor-fetch-order, asked Kaartic whether the client-side configuration is a sensible first step of the project. Shashank S posted a v3 proposal that dropped a pre-selected commitment to promisor.fetchOrder in favour of an early RFC, and Kaartic said the plan reads a lot better.