@prettygood It can never survive the dystopian future where every freely available search engine just gives out crap results. It worked good enough for me before, but the last few months were a complete train wreck of bad results, broken updates and almost daily rate limiting from search engines, because SearxNG just gave up and never showed any results even though the query was successful according to the "query report" thingy in the sidebar.
@eemmaa You in fact exist on multiple hard drives right now.
Embed this noticePhantasm (phnt@fluffytail.org)'s status on Wednesday, 04-Dec-2024 01:53:10 JST
PhantasmBought the Kagi $10 Professional plan just to try it out. I'm done with SearxNG for the time being as I just don't have the time nor interest to deal with badly interpreted search queries resulting in complete nonsense being returned back, dealing with crappy results from it's "meta" functionality that seems to be broken for months and only one search engine seems to be represented in the results (usually bing or ddg even though I also have brave and qwant enabled) and the frequent nonsense returned by ddg. A year ago it worked fine, but it seems that now every search engine is so polluted with shitty results that not even a supposedly "meta" search engine can sieve through the garbage to find something actually useful.
@sun@RustyCrab@SuperDicq@white_male@Nudhul It also had an alternate frame rendering mode that suffered from inconsistent frame delivery if each frame took different times to render. So you always traded either noticeable stripes in different positions on screen (because the frame split was dynamic) or you had inconsistent frame pacing.
@Nudhul@RustyCrab@SuperDicq@white_male GPU overclocking is basically dead anyway. nVidia sets the maximum clocks, voltage and power consumption and does not allow any change made to those maximums without resorting to external hardware modifications outside of VBIOS that the manufacturer will no be able to sell anyway due to the licensing.
In other words the "overclocked" cards are just cards with something better cooling than the stock FE cards with the default clocks slightly adjusted by the manufacturer. The BIOS itself will not allow anyone, even the manufacturer, to go above limits, which is why GPU overclocking has turned into maxing out the sliders in MSI Afterburn and that's it.
Even though manufacturers modify the default nVidia VBIOS and are allowed to change some parts of it, the maximum limits cannot be changed without breaking licensing.
@SuperDicq@VD15 When you market at it children that don't have thousand dollars in wallets, it suddenly starts to make sense from a predatory business perspective. As a kid you could afford a $60 monthly subscription, but not a one time $1K investment.
@sun Yeah, that has zero chance of getting actually implemented unless Mastodon somehow forces it on everyone. The plethora of AP extensions is already a mess similar to the one of XMPP and it looks like nobody has learned from it.
@sun I guess nobody thinking this would be a great idea has read lain's article about it from years ago.
As for the objection: The resulting design will very likely not be compatible with multiple AP implementations and dealing with keys/trusting them will be broken (like XMPP) or overlooked and broken (Matrix). Due to the nature of Fedi DMs, you are essentially dealing with groupchats. The final product will be OMEMO, but instead of XML as the base protocol, JSON will be used.
@lanodan >Lack of library management so you can end up with pacman libraries being missing?
That shouldn't normally happen. The library pacman uses to do most of the work (libalpm) is owned by the pacman package. I've seen AUR helpers break when used as "system updaters", because they aren't rebuilt against the new version of the library and the linker properly throws it's hands up when the helper tries to launch itself again and can't find the proper version shared library. I've never seen it fail this spectacularly and the only cause I can think of is that it only partially extracted the new pacman package or one of the dependencies (lack of free space, power outage, killed by user/system).
Btw sending ctrl-c when pacman is running any kind of package transaction (include pre/post hooks) is the fastest way to innocently kill an Arch system. The internal package database isn't updated properly, which will lead to pacman confusing package versions and your initcpio very likely missing.
>Package manager fails to update the signing keys? Urgh, this shouldn't happen.
I've seen this happen only when it couldn't verify the signature of the new keyring package based on the old keys it had. Which is a common problem if the system is left sitting for a month or two without updates. And probably the reason why pacman doesn't enforce signatures for local package installs.
@lanodan@phrawzty I've seen pacman break a few times when using AUR helpers, that's fair.
With Arch there's always some manual assembly required. Which to some extent is a good thing if you want to learn about how the whole Linux ecosystem works under the hood, but I understand it can be annoying.
@lanodan@phrawzty I've migrated my Arch install to Artix OpenRC couple months ago and honestly I can't complain at all. Everything that isn't already broken upstream of Arch works perfectly.
@prettygood@kim@eris@mold I'm starting to take the "don't update unless needed pill" for basically every computer I own. Too many things just break randomly these days. The only reason why I run my own branch of develop is to catch these bugs before they get released, because I care about Pleroma and it being better than it is.
>Milestone 2.7.1 >Changelog message not included in the 2.7.1 release notes >the prepared_one function change does not seem to be in the federator.ex file >prepared.ex does not exist
That's the extent I look into it and at least according to this it wasn't merged into the stable release.
@i@RustyCrab@cassidyclown@FrailLeaf@sun The "device-level" approach seems like an easily doable one. Generally speaking everybody has a Google account anyway and they already require an ID/passport to verify your age in EU and probably also in the US. And websites can tie that in with "Sign in with Google" functionality.
Of course this completely disregards vendor lock-in, but I think that's not something the Australian government cares about.