Anthony Martinez @ajmartinez@librem.one

@twrightsman @kyle in Qubes not even changing manually worked for me, but the USB dongle the headphones came with worked fine and I even did a Matrix call with a friend using it after I sorted my Synapse and COTURN issues.

@rbrown @kyle interesting. It didn’t do NOTHING but it did put a lot of line noise into my headphones rather than any actual audio.

@twrightsman @kyle I’ve not seen an issue with it powering off with the hub plugged in so much as I’ve seen it run out of battery that way. The hub appears to drop off (dom0 and sys-usb both see it coming and going with an annoying frequency) and when that happens charging also stops. My shutdowns seem to be related to that happening while I was already low on juice and doing CPU intensive tasks that happily drained the battery dead. I’ve stopped using the hub for now.

@kyle nice! My backup imported smoothly as well. My insert key doesn’t seem to work, nor does my audio jack, but otherwise most everything is as expected save for the USB-C hub shenanigans. If I get any free time this week I plan on live booting PureOS to see if those things work there or not.

@rbrown dom0 and sys-usb both see the hub disappear and reappear so I think it may just be some module instability. Will burn a live USB of PureOS and see what happens in that environment. My old laptop is also on QubesOS, and that’s really where I do most of the things where Qubes magic is really beneficial.

The audio jack doesn’t work at all in Qubes, but that was true of my T460s too. A USB audio device sorts that, and my headset came with one. The hub has one too. Both work.

SDKs that offer wildly different functionality in one language versus another really displease me. Lookin’ at you, Azure.

@absc@mastodon.uno @loweel most of the time my homeserver isn’t having huge problems, but when it does synapse is generally to blame. I do have much more machine, but it’s under-utilized. Still, I feel like most things ought to be able to run much lighter. I used run NAT with QoS, an MTA, and SIP on a 586 with 128MB of RAM and a 512MB disk. Shouldn’t need a quad core EPYC and 8GB of RAM to run a website and a chat service.

@loweel yeah, I was told not to run an instance on an embedded device... it’s a quad core EPYC with 8GB of RAM and NVMe block storage...

My server was created because I wanted to secure my own chat and voice with friends during my overseas assignment. When I did the same in 2005 during a combat deployment I setup Asterisk on an embedded device I hooked up to very slow (and expensive) satellite internet. That worked flawlessly. Guess this is what progress looks like 😂

It took two days of troubleshooting by way of trial and error in an absolute vacuum of useful information. Neither COTURN nor Matrix communities were what one might term useful in any way - with both often outright denying an issue was even possible. This isn't how you grow adoption.

I'll post my working config later, as well as some notes on how I'm dealing with Ubuntu's COTURN package having no good way of accessing Let's Encrypt certs thanks to an ancient version of systemd.

#FOSS misery

@kyle my Dovo may actually need to be sharpened after like a decade of getting by just honing on a leather strop. I have the stones here with me to do it. Maybe I will.

@kyle interesting, I've not used a safety razor of that type (vibrating or otherwise). Just my straight razor. Haven't used that in a while either if I'm honest.

Getting to the point where I will soon be someone who *used to* run a #matrix homeserver. The spec is a disaster, the reference implementation does not work, and it's increasingly difficult to justify asking anyone to use one of the many underwhelming clients just to send me a message that happens to be encrypted end to end.

For now I'm just going to keep using the built-in ports and stop messing with the USB-C PD hub and my Librem 14.

https://ajmartinez.com/tech/posts/202126-003-librem14

Added a very minimal "gallery" per the specifications of A Rust Site Engine's Product Owner (aka my bride to be) for the v0.10.1 release.

https://crates.io/crates/arse

#Rust #FOSS

@veer66 That's true, but unless you're running a daemon that interpreter startup time is per execution and the build cost (for a given version) happens exactly once. Of course I think this is amplified on simpler problems where the comparison is fairly simple std lib vs std lib work. Otherwise like most things it's about the right tool for the job, and the compile time certainly factors in!

@purism @kyle heh this same photo was used in an embedded security workshop I attended this week!

@veer66 I know the feeling. One of my Advent of Code solutions that disgusts a real software engineer for its inefficient algorithm still *finishes execution* before Python can even launch its interpreter. Still, Python remains a useful tool on which I have created quite a few mission critical tools.

@veer66 good deal. I generally turn to Python when my Rust chops aren’t enough. If Apache Arrow handles your use case rock on with it!

@rbrown bumping up the RAM in sys-usb didn’t help, and I bounced the VM without checking dmesg and friends. When the screen goes dead over USB-C it also takes out charging which I believe contributed to my dead machine symptoms. At present I am just not using the USB-C hub with my Librem 14. It has enough native ports to suit my needs, but I may try just a dumb hub with no more than PD pass through and USB3.1 ports to see how that goes. Charging itself has not been a problem directly on the port