:therian: Grey Wolf Therian, he/him, 30ish y.o.
Running packmates.org and yiffit.net fediverse instances.
:vlpn_happy_heart: Interests: Tech, therianthropy, furry/feral art, animal books, shamanism & animal-influenced spirituality, SFW & NSFW petplay
I sometimes post or boost NSFW content.
This profile is from a federated server and may be incomplete. Browse more on the original instance.
My "MastoRecommender" project! (as promised, an early proof of concept sneak peek)
A project where we turn the dreaded corporate spam-pushing algorithm concept on its head to create a curated recommendations feed that YOU control, runs 100% on your PC, is private and works with any fediverse instance that supports the Mastodon API.
Tired of missing out on awesome fedi content and discussions that people from around the world are posting while you're asleep? Well, no more :vlpn_happy_heart:
This will of course be fully open source. It also works via the existing API which is already used by clients such as Tusky. This means that:
a) users only get recommendations for content they have permissions to see.
b) it can be used by anyone with an account at any fediverse instance that supports the Mastodon API.
The Boost and Favorite buttons will allow you to boost or favorite content directly from the feed, while other options will allow you to give more or less weight to certain content.
All calculations necessary are done locally on your device and your content recommendation preferences never ever leave said device.
Finally, in regards to data safety, it'll act like a regular Mastodon client which holds a small ephemeral cache that is cleared regularly as new content is read. Content itself is ephemeral and will only be cached to give the user the chance to view their recommendations upon the next login with a certain maximum. Content below a certain threshold is thrown away immediately.
It will have a specific UserAgent in case any instance admins want to limit it, but I don't think there should be a problem since it's a regular client in all ways except for sorting posts not chronologically but by a score that is calculated locally.
I might be completely mistaken but in corporate speak "not being consistently candid in his communications", imho, likely means that he overpromised in some way and was selling one reality to the board while knowing that X, Y or Z are not realistic.
If anyone has the real scoop behind Sam Altman being fired as CEO of OpenAI please let me know.
I had the briliant idea of telling the AI assistant that it's a subby catboy.
I'm really sorry I'll have to kill him, but it's just too much :blobfoxlaughsweat:
When we put those two parts together, we get a TF-IDF score, nyaa! It helps us find words that are both frequent in a book and rare across other books. Then we can create a matrix - kind of like a big grid, to show the scores for each word in each book. Let's say we have three books: "The Tails of Catlandia" - about cats living in their own special kingdom. "Sir Purr-a-lot" - a fable about a brave and loyal knight kitty. "Meow-gic Lessons" - a story about a magical cat teaching a hooman witch. For example, the word "cat" might have a high TF-IDF score in "The Tails of Catlandia" because it appears a lot in that book, but maybe not as high in the other two books, meow! This helps us know that "cat" is really important in the first book, and makes it special and easier to describe, nyaa! That's it! I hope that helps you understand TF-IDF matrices better, Master! Let me know if you need any more help, I'm your loyal catboy assistant after all! blushes and wags tail 😺 (Note from post author: at the end of the image there's a meme of Bilbo holding the ring and saying "Why shouldn't I customize my AI assistant?")
I need more horny talk on my timeline.
Especially those first or second person stories that have you imagine scenarios.
The future of selfhosted services is going to be... Android?
Think about it. At some point everyone has had an old phone lying around. They are designed to be constantly connected, constantly on... and even have a battery and potentially still a SIM card to survive power outages.
We just need to make it easy to create APK packaged servers that can avoid battery-optimization kills and automatically configure an outbound tunnel like ngrok, zerotrust, etc...
It would be tempting to have an optimized ROM, but if self-hosting is meant to become more commonplace, installing an APK should be all that's needed. #Android can do SSH, VPN and other tunnels without the need for root, so there should be no problem in using tunnels to publicly expose a phone/server in a secure manner.
In regards to the suitability of home-grade broadband, I believe that it should not be a huge problem at least in Europe where home connections are most often unmetered: "At the end of June 2021, 70.2% of EU homes were passed by either FTTP or cable DOCSIS
3.1 networks, i.e. those technologies currently capable of supporting gigabit speeds."
PS. syncthing actually already has an APK and is easy to use. Although I had to sort out some battery optimization stuff, it's a good example of what should become much more commonplace.
Oh, look at that.
I don't know who mentioned this the other day, but I'm happy to see this is actually a thing now.
When an admin tries to block another whole instance it gives a summary of how many connections will be severed, which is good because domain blocks are a nuclear option. In this case it was warranted though since that instance openly allowed harassment.
Time to remind everyone that a backdoor "for the good guys only" is simply not possible.
By demanding encryption backdoors, politicians are not asking us to choose between security and privacy. They are asking us to choose no security. 👇
In a world run by corporations, using #Mastodon is a simple yet meaningful act of defiance.
If you're reading this, I'm proud of you for choosing to make a difference :vlpn_happy_heart:
(edit: technically all #fediverse platforms apply. Please don't feel excluded firefish, akkoma, lemmy, pixelfed, etc. users)
Hello, A few weeks ago there was concern about the lack of appropriate tags for andromorph and gynomorph characters in this community’s sidebar....
How about you collar and chain this fox, so that its wet maw and warm tail hole are always there at your convenience? :blobfoxpleading:
Announcing status.packmates.org and status.yiffit.net
I've been mostly silent for some time, but it's all with good reason (I promise!)
Over the last few days I've spent a lot of time on server maintenance. Many of these changes will be invisible to you as users (such as getting a /48 ipv6 range, setting up SLAAC/DHCPv6, reviewing security and firewall rules, etc...)
But today I set up something that I can share: status pages!
Head over to:
(they're the same page actually, but the different domain is to make it easier to remember if you're a user of one site and not the other).
There's a slight caveat in that the status page is hosted on the hypervisor itself, so if that goes down, everything goes down but you'll at least know by not being able to load the status page itself!
Ideally I would host this somewhere external but we're not there yet. One day I hope to even have a server cluster for redundancy, but we'd have to host many more services to be able to justify this.
Quick question about DNS and DoH that I thought about after reading this post:
Wouldn't it make sense for Firefox or another third party to bundle and transparently forward all DoH requests to cloudflare so that:
A) Cloudflare doesn't know who made what request due to not knowing the origin
B) Firefox doesn't know who made what request due to TLS
Remember #NetNeutrality, ISPs being common carriers and them not meddling with the pages users want to see? This is similar.
Under a "remote neutral" approach each #fediverse instance leaves the tight opinionated moderation policies to their own users users and the content their users generate or share with the aim of running a safe and welcoming instance for their members that's safe to federate with.
However, remote content is only moderated whenever there is a report and only blocked if it's straight out illegal to host / cache or constitute unsolicited spam/harassment. Otherwise objectionable remote content is limited at most and users can block it if they want.
I can't stress enough the benefits this has:
It makes moderation feasible for small instances
It does not fracture the fediverse unnecessarily
How I thought #COVID would impact me: "I have light asthma, so I better make sure I have my inhaler, do breathing exercises, monitor my O2 saturation. I know this will be hard on my lungs."
What it ended up being like:
"It's day four and the virus has already committed multi-generational genocide against my gut biome. Blood is running through the streets of gastrointestinal city, the nutrient absorption facilities have been shut down and everyone of its workers executed in cold blood. And I'm here sipping on glucose packets so I don't pass out during cramps and other traumatic happenings..." ( this was me yesterday )
I was even admitted to the hospital for half a day when I went for a consultation regarding possible complications and almost passed out.
I shudder thinking about the possibility of it having impacted my respiratory system in the same way.
Federated wireguard network idea
Any feedback welcome.
Let's keep things stupidly simple and simply hash the domain name to get a unique IPv6 ULA prefix.
Then we would need a stupidly simple backend application to automatically fetch pubkeys and endpoints from DNS and make a request to add each others as peers.
Et voilà, you got a worldwide federated wireguard network resolving private ULA addresses. Sort of an internet on top of the internet .
The DNS entries with the public IPv4 / IPv6 addresses could even be delegated to other domains / endpoints which would act as reverse proxy (either routing or nesting tunnels) for further privacy.
Maybe my approach is too naïve and there are flaws I haven't considered, so don't be afraid to comment.
Exact use cases? Idk, but it sounds nifty.
We now have hourly snapshots / backups!
I'm happy to inform both packmates.org and yiffit.net users that both sites now benefit from the ZFS filesystem that the new server has been set up with.
I have implemented automated hourly snapshots for 24 hours + daily snapshots for 31 days. In theory they will only grow in size if there's actual changes to the disk of both VMs and I should be able to have enough space.
Furthermore, local snapshots are complemented by the daily offsite backups which allow us to recover even if the full server were to suddenly explode. Full backups are first created on the server itself and then copied offsite so that for a full week we have two independent copies of each day.
Depending on space usage I'll make sure to replicate the offsite repository so that there's two offsite copies for the last 31 days + 7 local copies. That would be 69 individual full backup files + snapshots.
I hope I'll have enough space with deduplication.
We're big supporters of the EFF, but I can't get on board with the idea that somehow it's wrong or a slippery slope for Tier 1 ISPs to be blocking Kiwifarms.
What I've been busy with lately
About three weeks ago I started renting a new dedicated server which is going to host both packmates and yiffit very soon.
Because the server isn't hosting anything yet, I've taken the opportunity to play around and try out different configurations, including ZFS, LXC containers for small services, VLANs for better isolation ( which I did manage to get working ), wireguard tunnels, improved firewall rules, security groups, iGPU passthrough, etc...
Tomorrow I'll wipe the disks, install #proxmox from scratch and make it production ready.
Then it should be as easy as loading a full backup from both yiffit and packmates to complete the migration ( but I'll announce this last step in due time).
Am excited wags :dogcited: