Linux VFAT question
Are there cases where Linux can return a different inode number for the same file even if the filesystem hasn't been unmounted and remounted? I'm seeing a strange issue with git-annex that Joey says would have to be caused by the inode number's changing between listing the files and fetching their content. This is an SD card with 160k+ files on it.
The version of #gitannex in #Debian Buster and buster-backports has a bug that causes it to leave files open, at least when importing from a directory remote on VFAT, so if you are trying to import from a directory remote with a lot of files, it can run out of file descriptors. Latest git version does not have this problem, so just build from source. Everyone has a #Haskell toolchain installed, right?
Anyone have thoughts on how much processing power is required for cycle-accurate emulation of a #Commodore 1541 floppy disk? Would a 48Mhz ARM Cortex-M4 be sufficient? I like my #Pi1541, but I would like something between it and the uIEC: full support for browse mode and regular listing of directories (you can't 'load "$"' with Pi1541 in browse mode), simple on-board UI for swapping disks, and cycle-accurate emulation, but no USB or HDMI and on a single board.
Anyway I ordered one of these, a well-ventilated case with a DC-DC converter, and an SSD in hopes of building a completely silent Plex server and testing out a compute node candidate for a Kubernetes cluster. Perhaps not ideal since it only goes to 8GB and requires SODIMMs, but it's the cheapest way I've found to achieve those specs.
Any thoughts on why X freezes periodically on my 7th generation Thinkpad Carbon X1? It's running Arch Linux, and this has happened across multiple kernels. It happened before and after I updated the firmware to the latest with fwupd. The common factor seems to be Firefox, though it's happened across multiple versions and for a long time. It most commonly freezes while I'm scrolling Facebook which makes me think it might be RAM usage related. I'll stop using FB to see if that helps.
Holy crap, the Faraday FE2010A for the Retrobrew/Xi PC is *expensive*! $30 a piece from Kynix + $40 shipping, or $50 a piece from Monotech Vintage PCs, plus $11 shipping to the US. I bought two from Monotech since I figure at least one person will be interested in a kit, but if I wanted more, Kynix would have been cheaper. Maybe I should have been more patient and tried to get together a group buy.
There are all kinds of ways this could be abused. All it takes is for someone to actually have the motivation and knowledge to abuse it. You can protect yourself in private somewhat by having the people with you turn off their devices (assuming they actually turn off). You can't protect yourself in public, though there is less expectation of privacy in public.
I fear nothing will get the general public to care about this stuff short of someone actually exploiting it and causing harm.
Saw an ad for "Echo Frames" the other day. These are glasses with open ear headphones and a microphone. It got me thinking about the number of devices around us that are listening all the time. They don't send all of the audio to a server; that would use too much power and bandwidth. Instead they analyze the audio locally, ostensibly just looking for a "trigger word", then send the surrounding audio, including some from before it thinks there was a trigger word, to Amazon's servers.
I think my insistence on #retrocomputing "purity" has actually been harming my motivation to play with my #RC2014 because it's such a pain to connect to it. Even when I get the keyboard driver finished, the Dupont wires limit where I can have the computer and keyboard, and the Universal Micro Keyboard won't be great to type on. Maybe I should just set up the C64 as a terminal?
The interface board is at https://hackaday.io/project/167449-zx81-keyboard-adapter-for-rc2014 (I have extras from my JLCPCB order if you're in the US) and the actual keyboard I have is https://rc2014.co.uk/full-kits/universal-micro-keyboard/ . I haven't actually tested them together yet because I haven't finished porting the driver nor have I tried compiling Fuzix with an updated keymap (or tried Fuzix at all for that matter).
"The new group of at least 227 engineers and other workers is notable in the usually union-averse white color workforce Silicon Valley."
"white color workforce" sounds reasonably accurate, but I'm not sure it is what they intended to write.
Posting this with public visibility was a mistake.
I still strongly believe in decentralization. I don't like being stuck using Facebook. But I'm not really using Facebook *instead* of the Fediverse; even during periods where I don't really use Facebook, I don't start posting here instead. It's more that the novelty has worn off, and I just don't really think of this place anymore when I want to read or write something.
Another issue with the streaming nature of the Fediverse is that there's not a strong distinction between posts and comments, which makes "threadjacking" common. There's also the issue of messages in threads that the poster can't see. And the actual threading of posts and replies is hard to visualize or understand, and it's separate from who can see or is notified of the reply, with the result that I occasionally get asked to remove someone from a thread here.
I was skeptical of their "engagement" metric, but he wasn't talking about engagement; he was talking about people even reading posts or using Facebook *at all*, not just whether they liked or commented or whatever.
(Speaking of engagement, the reason for the lack of a "dislike" button was because nobody could say what "dislike" actually meant - but they discovered that if someone commented *without* liking a post, it was usually because the post made them mad.)
I actually worked on the newsfeed at Facebook back in 2014 & 2015. I asked the director of the product engineering team why the feed order setting kept reverting to "most relevant", and his reply was interesting, because it was backed by actual data: when people set their feed to chronological order, they start coming back less and less and eventually stop using Facebook. If they revert the setting before that happens, people immediately start using Facebook more.
But I think the worse problem is the chronological ordering of the timeline. If someone doesn't happen to look close to when I make a post, they probably won't see it. Likewise, I only see whatever was posted close enough to when I happen to look. And to be honest even though I tend to unfollow people if their posts aren't interesting enough to me, I haven't been able to get the signal to noise ratio of my feed high enough to make it worth reading if I'm not super bored.
I think the real difference is that this is just not a great medium to have deep discussions in. The 500 character limit is too short, and even without it, it's painful to type out and read long posts in most frontends. Even if I found a better frontend or a different instance, others would still be trying to read my posts in a narrow box.