Repository navigation
Replies: 1 comment
|
Carrying forward from #4680, since the comments stayed on the closed issue, plus what has shipped since: Part of this is solved, thank you. 0.60.162.0 (11 Oct) lists "Cross-source links the source policy approved are written by every writer (#4680)": What still seems open is the explicit verb. On 0.60.472.0, The ask stays the same, now narrower: let |
Uh oh!
There was an error while loading. Please reload this page.
Moved from #4680, opened by @knicolov on 2026-08-28. Feature requests now live in Discussions > Ideas.
Version: 0.47.3.0 · Engine: postgres
Observed
add_linkfrom a page in source A to a page in federated source B is refused:The suggestion offered is to use a client owning the target source, or to import the target page into the write source first.
Why the design is right and still leaves a gap
Not writing edges into a repo you do not own is correct, and should stay. But the refused operation was not that. It was: record, in my own repo, that my page references theirs.
That edge belongs entirely to the writer's source. Nothing needs to be written to source B.
Impact
Federation currently provides a unified read and search surface over a fragmented graph. Retrieval spans sources; the graph cannot. On a brain with five sources, the highest-value connections are exactly the ones between separately-acquired bodies of work, and those are the ones that cannot be expressed.
Concrete shape of the problem, generalised: a specification held in source B cites, by name, a paper whose full text is held in source A. That citation is a fact present in text already in the brain, and it is unrepresentable as an edge.
The two workarounds offered both fail. Running a second client per source does not scale past a handful of edges. Importing the target duplicates primary text, which is the defect the source separation exists to prevent.
Ask
Allow an outbound edge that is stored in and owned by the writer's source and points at a foreign slug:
Also worth documenting either way
Whether a local
gbrain captureingest resolves body wikilinks that cross source boundaries, or whether the same-source rule applies there too. That determines whether authoring cross-source correspondences as prose with wikilinks is a viable path or a dead end. I could not find this stated anywhere.All reactions