> They appear to be designed without pedaling in mind, as exerting effort without proper ergonomics would quickly become uncomfortable and painful. You can actually see some such bikes in the linked article.
Came here to write exactly that. Those who design those bikes clearly don't know a thing about bicycle design. Want to use them for pedaling? Say hi to knee problems and inefficient pedaling!
Yeah I was onboard until the re-encoding part: yt-dlp maintains the exact bits, why on earth would someone want to waste encoding time just to trash the quality?
On top of that... seriously, of all the formats one could choose, MOV?! Might as well choose DivX or RealVideo.
the video has to be re-encoded because apple quicktime doesn't like the youtube video format. But the audio can just be copied. My mac's fan never spins with the hardware acceleration so it runs in the background and I just forget about it.
> the video has to be re-encoded because apple quicktime doesn't like the youtube video format.
That’s not true at all. QuickTime is far from the best video player, but it’s also not entirely worthless. It can play “modern” popular formats like H264 MP4, which is exactly what YouTube recommends.
You can use the -F flag to list all of the available formats, then --merge-output-format mp4 of audio and video formats that will work best for Quicktime at the desired resolution.
Thanks. I played around with your idea and got this. It’s still not 100% of videos. Only YouTube videos that have any H.264+AAC stream available (which is 99.9% of public YouTube today, even if the main/default version is VP9 or AV1).
But re encoding to solve this is not required. I stand corrected.
Seems likely the reason is they want to easily load videos into mobile Apple devices like an iPhone or iPad. While alternative video players exist for those platforms, management may not be as convenient.
Why does Apple take the effort to maintain and ship different encoding libraries? I would've expected to both the Safari engine and Quicktime to simply depend on libappleavsmth.dylib?
That being said, that makes zero sense. Just linking to a library, doesn't precluded using a protocol over a socket to talk to a graphic/audio server. Access control like remote code isolation (webAPIs), CORS and DRM also don't change anything about decoding and mixing video streams.
Wait... 60 dollars per year for a third party solution just to send articles to your ereader? I'd argue that's something that should come with the reader.
On the GrapheneOS forum you will see a lot of bad opinions about F-Droid, for example this:
> It doesn't matter that the app is trustworthy, because F-Droid are extremely incompetent with security and the apps you install from F-Droid are signed by F-Droid rather than the developer.
> If the app is only available on F-Droid / third party F-Droid repo, use F-Droid Basic and use the third party repo rather than the main repo if available.
>
> If the app is available on Github then install the APK first from Github then auto-update it using Obtanium. Be sure to check the hash using AppVerifier which can be installed from Accrescent (available on the GrapheneOS app store).
By the way, while GrapheneOS recommends Accrescent, I don't use it anymore because they can't even add apps like CoMaps, while some of the apps they actually added are proprietary.
>the apps you install from F-Droid are signed by F-Droid rather than the developer.
That doesn't seem like a con if you take into account the context: F-droid is not shipping pre-build binaries from the developper, it asks for a buildable project from the developper.
If the source repo of the upstream dev are compromised, so will be hid own binaries anyway.
> [A]pps you install from F-Droid are signed by F-Droid rather than the developer.
Having recently gone through the F-Droid release process, I learned that this is not necessarily the case anymore.
F-Droid implements the reproducible builds concept. They re-build the developer's app, compare the resulting binary sans signature block, and if it matches they distribute the developer-signed binary instead of their re-built binary.
This is opt-in for developers so not all apps do it this way. I'd sure like to know how common this is, I wonder if there are any statistics.
F-Droid only uses reproducible builds for a tiny portion of apps, and there are still significant disadvantages. It depends on the app developers always complying with F-Droid's rules otherwise users are left without updates. F-Droid only checks that the build matches, they do not review/audit the apps and will not catch hidden malicious behavior or simply non-compliance with their rules. WireGuard's app deliberately broke F-Droid's rules by including a self-updater which was not noticed by F-Droid and shipped by F-Droid. WireGuard used this to start taking over updates for itself to migrate their users away from F-Droid. F-Droid eventually found out when the WireGuard developer brought it up many months later and couldn't do anything beyond dropping the app. It had taken over updates for itself already and F-Droid wasn't in the picture anymore.
The process adds a significant delay for updates but it does not actually protect users from developers in any meaningful way. This real world example with WireGuard demonstrates that.
> Work profiles are inferior to separate user profiles, which are built-in to GrapheneOS.
Different use cases. User profiles are only active when you manually switch to them, while work profiles are active _alongside_ your main profile.
So for untrusted apps that you only use occasionally and on-demand (like the myriads of travel / shopping / random services apps), user profiles are great. For apps that you want to keep in the background, such as the proprietary messaging apps that all your friends use, a work profile is much nicer.
Private Space is very similar to a user profile but nested inside of another user. GrapheneOS adds shared clipboard control for Private Space which was the main disadvantage compared to a secondary user.
GrapheneOS supports having a Private Space in secondary users instead of only a single one in Owner. Supporting multiple Private Spaces per user is a planned feature at which point work profiles will be fully obsolete. The remaining use case for work profiles is to have both a Private Space and work profile in the Owner user.
Don't you have user profiles in Pixels? I can create another user an switch. Just not super convient.
Work profiles are actually pretty good good... For work.
Just to add to that: Even some proprietary applications let you download their APK right from the website. WhatsApp is one such example (I don't recommend that you use it, Signal is much better, but if you require it, you don't have to use the Play Store).
Private space is identical to work profile. In the past, private space didn't exist and people used work profile instead as a workaround, but now that's not needed.
Private Space has a superior approach to isolation and encryption matching user profiles. Work profiles have some compromises for historical reasons. Private Space should be preferred over a work profile and the only reason to use a work profile for your own local usage is to use both a work profile and Private Space at the same time. Once GrapheneOS has support for multiple Private Spaces within a user, the use case for work profiles will be limited to the intended Bring Your Own Device enterprise deployment purpose. The intended purpose of work profiles is companies not having to give their employees work phones but rather owning/controlling a specific profile on their device with some influence over the overall device via rules for lock method, etc.
But you can't do that on a live system as you can with Windows or macOS. Not a problem for pre release upgrade perhaps. But I'm so missing this feature from macOS.
You can if you're using LVM. Take a snapshot of the logical volume your system is on, then run `dd' against the snapshot, as it's essentially a frozen point-in-time.
I've used this trick many times in a live, rw environment.
Depends on your filesystem. For example, I certainly can as I’m using btrfs. I’m also using Timeshift for easy management of snapshots. As others have mentioned, there are other choices too like Snapper that all work well.
> According to Google reCAPTCHA, Hcaptcha and Cloudflare Turnstile, I’m a robot. Although I haven’t seen an engraved plate with the label T-800 on myself yet, these days I might as well just be a rogue LLM. Does it even matter anymore?
reply