Congrats on solving your issue, and thanks for updated the post with the solution! It's really slick how NixOS makes adding a kernel patch to your config no more complex than it would take patching any dot file. Hope the up streaming of your device info goes smoothly.

You could get a fiber optic display/HDMI cable, a fiber optic USB cable, and the USB hub, then just move the desktop tower into another room and run the cables through the walls or ceilings to your display setup. Might only be $100 or so cheaper than then a used business thin client, but at least you could still do something 4K 120Hz HDR 12bit over some distance without compromise. E.g:

• https://www.youtube.com/watch?v=lsuE0YjfTNI
• https://www.youtube.com/watch?v=BMn6PvhL_Ms

A while back, I tried looking into what it would take to modify Android to disable Bluetooth microphones for wireless headsets, allowing for call audio to be streamed via regular AAC or aptX, and for the call microphone to be captured from the phones internal mic. This would prevent the bit rate for call audio in microphone being effectively halved when using the ancient HFP/HSP Bluetooth codecs, instead allowing for the same call quality as when using a wired headset. This would help when multitasking with different audio sources, such as listening to music while hanging out on discord, without the music being distorted from the lower bit rate of HFP/HSP. This would also benefit regular VoLTE, as the regular call audio quality already exceeds that of legacy Bluetooth headset profiles.

Although, I didn't manage to tease apart the mechanics of the audio policy configuration files used by the source Android project, given the sparse documentation and vague commit history.

• https://source.android.com/docs/core/audio/implement-policy
• https://android.googlesource.com/platform/frameworks/av/+/dc46286/services/audiopolicy/config/audio_policy_configuration.xml#147

I'd certainly be fine with the awkwardness of holding up and speaking to my phone as if it was in speaker mode, but listening to the call over wireless headphones, in order to improve or double the audio quality. Always wondered what these audio policies fall back to when a Bluetooth device doesn't have a headset profile, but it's almost impossible to find high quality consumer grade Bluetooth headphones without a microphone nowadays.

For the call setting under Bluetooth audio devices, I really wish they would break out or separate the settings for using the audio device as a source or sink for call audio. Sort of like how you can disable HSP/HSF Bluetooth profiles for audio devices in Linux or Windows.

Has Bryan done any more recent recorded talks?

The only experience I have with working with Fortran would be setting up gfortran when building SciPy from source, and perusing its codebase to see how it's FFT functions were so optimized. Not enough to diligently mod I'm afraid.

I'm not sure why, but GitHub's search engine, Blackbird, seems to be returning some erroneous results for this query:

• /tnt_select\(.*2\^32/ language:C++ OR language:C
  • https://github.com/search?q=%2Ftnt_select%5C%28.*2%5C%5E32%2F+language%3AC%2B%2B+OR+language%3AC+&type=code&ref=advsearch

Any chance you could narrow down your search to a list of repose that use the library that pulls in tnt_select() function, then clone and manually grep just those, or is it's use too common to index by?

Real funny that even narrowing down GitHub search to just the same repo doesn't help the query results:

• repo:ocelot-inc/ocelotgui ldbms_tnt_select
  • https://github.com/search?q=repo%3Aocelot-inc%2Focelotgui%20ldbms_tnt_select&type=code

That was really cool and got me inspired! Thanks for cross posting.

Yeah, I suspect the current poll rankings are a little biased from the initial ordering of the ranked options. I think strawpoll.com's website should randomize the initial ordering per voter, as well as ask the user to move options from one column to a second columb to make the UI more apparent that all options should be re-ranked by the voter.

When @'ing or pinging someone by username, the auto complete will suggest matching users from across federated instances. So typo squatting could be a thing, or user impersonation. But it'd be rather transparent when it occurs, but some could still fall for such tactics, just like with email spam or phishing scams.

If folks vote for the UBP option, then I think that unified color palette and aesthetic would be a reasonable enough clue of instance origin.

That said, if we did hardcode shapes into our instance icons, it's not like there's a lot of maskable radial shapes to choose from that would:

• fit inside the default circle that remote Lemmy instances would render our community icons as.
• not overly crop out or limit the artwork canvas with fat dilated margins.
• https://maskable.app/

With hexagons, octagons, rotated 45 or 30 degree offsets, at tens of pixels font heights: those are hardly differentiable at some point. So regardless, we could easily collide with other servers already that choose hexagons, or any polygon that fits in Lemmy's minimum circle mask.