Hardware survey/telemetry

Does fedora have any sort of hardware survey or telemetry? I think anonymous opt-in hardware survey would be really useful. Especially for conversations like F40 Change Proposal: Optimized Binaries for the AMD64 Architecture (System-Wide) - #35 by amoloney . How do you even go about discussing this if you don’t even know how many people would benefit?


If such a thing doesn’t exist, could add something to /etc/skel so first time someone logs in they get a pop up explaining what information is collected and they have to check a box to opt in.

A pretty anon way to generate an id to prevent duplication would be cpu serial # sent through sha512 or sha256. No way to reverse it but then the same person can’t send it twice.

This came up on the fedora devel list a few months ago and went no where
I recall these sentiments comeing up:

You cannot succesfully anonymise the data sent.
When asked to opt in with a default of don’t there is an expection that few replies will be sent.
Defaulting to sending is seen as a dark-pattern and very much disliked as an idea.

Large deployments such as student labs and cubicle farms have only a few admins who interact with the Fedora services but many users. That makes it difficult to really understand the characteristics of the full set of Fedora users. Vendors selling systems configured with Fedora will have better understanding of their customer base. Such vendors sometimes provide drivers and firmware (e.g., for recently introduced chips) that eventually get accepted by kernel.org. Such vendors could start delivering AMD64 optimized linux, and patches should eventually arrive in kernel.org.

large deployments are part of the userbase. It’s not really skewing it is it? I wonder how many large deployements use fedora and not rhel or a centos derivative. That would be a good question to answer with telemetry. Although this idea would not collect that information.

There is https://linux-hardware.org/ although it seems to be down at the moment.

Linux gets features and configuration patterns driven by the people/entities who need them, but it can difficult to predict collateral damage to existing user-cases. We have seen issues, e.g., default Fedora Workstation power saving settings driven by vendors’ need to meet energy efficiency standards, that aren’t appropriate for users who run services on Fedora Workstation. We could see vendors providing pre-installed linux with binaries optimized for their hardware if they see an advantage. I think a linux distro should be careful about introducing innovations that don’t have upstream support. Where work is needed is finding ways to avoid breaking existing use cases and reducing barriers to users new to linux.

That isn’t really relevant.

It seems like you’re trying to come up with reasons to not have a hardware survey but in doing so actually giving more reason why one is needed. If an appropriate hardware survey were done, then they could have known in advance not to enable such a thing or if it should be.

No, I’m concerned that some important (to vendors and something we hope will become more common over time) deployments could be missed because they target a class of users that rarely if ever touch Fedora servers because the deployments are done by their institutions.