mranderson17

I use sway and run zoom in my browser (because zoom is shady and I don't trust them). Screen sharing works fine in the browser. The application never worked very well to being with anyway for me, even on X11.

I also use https://git.dec05eba.com/gpu-screen-recorder/about/ for individual output screen recording such as gaming which works amazingly well. You can not select a section of a single output though, only the whole output. That's a deal breaker for some, and a non-issue for others, just depends on what you need.

I think this one https://github.com/nextcloud/desktop/issues/5369 is probably the more relevant, and also open, issue. However even in that issue people claim you can choose not to. The argument is only that it suggests restarting explorer and also rebooting and that this is annoying. So you never get a prompt, it just dies?

I agree though that the amount of time where it was force rebooting is pretty bad, and it looks like the rollout of the patch was mishandled. I also should probably admit that I've never touched the windows client, my environment is entirely Linux and Android. The Linux client even with file manager integration doesn't require restarts of anything.

I mentioned the client in there (4th paragraph), but mine was more of a general rant on the overall low effort that seems to have been put in to figuring out what the actual problem was. And that it is relatively common among people in the self hosting community to assume that Nextcloud is a lot simpler than it is. It's a huge cloud suite consisting of many applications, clients, plugins, proxies, caching, database, etc. You need to have a pretty good understanding of how it all works, and how to investigate a problem, and ideally you should be testing before upgrades. Large organizations often even test endpoint applications like the desktop client and push out only tested versions to users via policy or some kind of endpoint management.

I can't really draw many conclusions from the very little information provided in this post, but I suspect OPs windows machine is not in an entirely stable state, which is what is causing some of these update issues.

And, I put some of the blame for Nextcloud under-representing it's complexity on Nextcloud's marketing and AIO. You absolutely can install it without understanding anything, and that's a little dangerous in my opinion because it is actually quite complex and you will probably end up breaking it at some point and need to dig in to fix it.

Ok, I'm prepared to be downvoted today so here goes.

Nextcloud is an enterprise cloud suite. The one you run in docker on your rpi (or whatever) is the same one that is run at a company, albeit with more high availability and redundancy, but the same application, proxies, caching, db, etc. Nothing is stopping you from running the stable channel and testing your upgrades, or even rolling out specific stable client versions to your devices.

Said companies often have teams (more than one person) to run it, stage upgrades, automated testing, automated backups, monitoring, etc. They go to work and do just that, maybe not every day but at least a couple times a week their focus is Nextcloud and only Nextcloud.

What many people in the self hosting community do is spin up docker, without ever having touched docker before, and try to run Nextcloud, forget that it exists, and then upgrade it a year later across multiple versions without maintaining the database. Then they obsess about how fast an app loads by refreshing it a whole bunch, and then complain on internet forums that it sucks. This, like many posts, doesn't have a specific problem for us to help with, no logs or stack traces have been posted, and the subject of the complaint shows just how terrible your understanding of application security is.

So, while there is legitimate criticism of some of Nextcloud's design choices, this isn't it. And at the risk of sounding a little gatekeepy, if you post "nextcloud updates break everything" with no context you probably should spend some time gaining a better understanding of how internet facing services work and make an attempt to fix the problem (probably misconfiguration, and in this desktop client case probably a heap of un-updated local software installed alongside the client), which I'm sure people would find if they did the bare minimum of reading a few log files or any of the other things that come with being an application admin.

“An attacker would need to be able to coerce a system into booting from HTTP if it's not already doing so, and either be in a position to run the HTTP server in question or MITM traffic to it,” - Matthew Garrett

Summary left out a quite important bit.

Mosh hasn't had a release in quite a while (Oct 2022). While that's not that old, and there does appear to be somewhat active development, it's a little slow moving for something that might be open to the internet directly. I used to use it but ssh with tmux is mostly fine and makes me feel a little safer because of their wider use.

AFAIK openSCAD is a code driven mesh format. So if you want to import openSCAD models into any other CAD software you have to convert the mesh to STEP or some other actual 3d object format during which there can be lots of error if the model is complex. I don't have a lot of experience doing this but I just tried a model I had lying around from the dactyl keyboard project and converting it resulted in a lot of really broken surfaces.

This is a cool alternative that makes 3d objects instead of meshes (at least it says it does). https://zalo.github.io/CascadeStudio/ . Also open source but web based.

EDIT: I should mention that CascadeStudio seems to be abandoned, just a cool concept of a different way of doing code driven CAD.

I use FreeCAD and Assembly3 for everything and have for many years now. I sometimes use realthunder's fork of FreeCAD but right now it's quite a bit behind upstream and there are some cool new features in sketcher so I use upstream for those.

Some people get confused about workflow in FreeCAD because there are so many options and every youtube video has different opinions or tries to feature a particular workbench like curves or something. My opinion.... Pretty much your workflow starting out should be to ignore everything else and use part design and sketches, it's the simplest way:

0. enable autosave with a short interval, like 2min

1. Switch to part design workbench

2. create body

3. create sketches as the base of the features of your part attached to the xy, xz, yz planes, offset them to create a "wire frame" that resembles your project

   a. Your sketches should be fully constrained

   b. Your sketches should have as little geometry in them as possible, if you need more complex stuff make more sketches

   c. Your sketches should have closed wires, you can't pad something that doesn't create a face.

4. use pad, pocket, revolution, loft, and hole operations on those sketches to form a 3d solid

5. if you need to create additional sketches which import geometry from the previous operations (using the external geometry tool), import SKETCH geometry from the previous ops, not edges of solids, whenever possible. Hide your solid, unhide your sketch, select that with the external geometry tool.

   a. Use sketch on face sparingly.

6. Do fillets and chamfers last, if you need to change something, delete them and recreate them once you've made your changes.

To make multiple parts make multiple bodies with the same workflow as above.

Once you get pretty good at making static parts with constrained geometry, holes, threads (with the hole function), etc, which you can do with only the stuff above, then you can branch out into other workbenches like assemblies or curves, but all of those things build on the concepts above, so it's easy to get overwhelmed if you try to do it all right from the start. Learning how to recover from a mistake is just part of CAD in general, though I admit that it's a bit more effort to find what's wrong in FC vs commercial platforms, but we aren't here, on lemmy, in a linux community, to use commercial platforms.

AFAIK that's pretty much the same workflow as F360 uses for single-solid parts though things have different names. pad=extrude for example.

It's obviously far from perfect but in my opinion it's the best solution that runs natively on Linux and is actually open source. Also assembly3 uses solvespace as it's backend solver so if you make assemblies using that you are kindof using solvespace too.

Also, I hear/read a lot of complaining about instability but I've honestly never had a crash that wasn't on an experimental branch like RT or the edge release of upstream. However step 0 above should help if you're worried about that.

this one? =]

I do this too, but additionally group these outputs strategically on my 4 displays. I never thought of it like a desk with papers on it but that's very much what it is. And also how I organize papers on the few occasions that I do that.