filter hashes that are unlikely to be useful

[raboof]

Some hashes will be unlikely to be useful to share, as they are specific to users' configurations.

Could we somehow identify those and avoid uploading them? Or perhaps periodically garbage-collect 'old' hashes that only appear once?

[JulienMalka]

I was thinking we should evaluate nixpkgs revisions that hydra evaluates, or parse hydra evaluation results in order to only accept derivations from nixpkgs. This is also a matter of privacy because currently we are also storing derivation names, which can carry information.

[raboof]

I was thinking we should evaluate nixpkgs revisions that hydra evaluates, or parse hydra evaluation results in order to only accept derivations from nixpkgs. This is also a matter of privacy because currently we are also storing derivation names, which can carry information.

That seems like a good start to me.

If we implement that on the server side, there could be a race between rebuilders and #6 (or whatever process we use to determine which derivations are from nixpkgs), so perhaps we should do this in the build-hook at least initially?

[raboof]

Maybe as a first first step we should give the build-hook a 'whitelist' parameter that we can use to only upload builds that are part of a report.