• 0 posts
  • 10 comments
Joined 1 year ago
Cake day: July 3rd, 2025
  • The content is going to be both text-based and media like videos pictures etc. Currently we are on a telegram super group but that doesn’t cut it and can be lost. Is piefed or Lemmy a good fit for this? I have been here since 2023 and I think it would be. One other issue is privacy the identity of the members should be protected at all costs.

    If this is only for having a place to socialize, Fedi will work

    But if you want privacy, maybe check out delta chat? https://chatmail.at/doc/relay/getting_started.html
    You probably will still need a domain for the relay and somewhere to run it but when used in one account mode, delta relay stores nothing on users. It only relays the messages

  • I have also been wondering about this. But

    admins and mods are selected sortitionally( randomly) from people who apply

    That’s your point of starvation. The application

    I think, instead, everyone on the instance should be a voting member. And the decisions should be taken based on some requirement on the votes

    1. person one flags a post -> that’s a vote
    2. maybe add a header that this post is under review? The voting process should probably be taking some time
    3. another person flags the same post -> that’s a vote
    4. someone else says it should not be flagged -> that’s a vote in opposite direction
    5. ??? <- can be changed by an instance wide vote too
    6. post/comment/user is either moderated or not, depending on votes and logic from the previous vote

    In order to boost participation, randomly ask an active account to cast a vote on an ongoing discussion

    I think there should be some way for “backend admin” to step in and ban something. But they should only do that for spam/csam/scam/etc. Nothing else, no matter how crazy or rude the take is
    (btw, maybe the instance could also normalize asking someone to reword their take if it’s rude or can be interpreted badly out of context?)

  • This is dangerous

    Hence my point about why bother at all. Without full encryption one gets leaks. With full encryption some kind of secret is required. Either password (hence that need of keyboard in earlier comment) or a key, etc

    In order to not need a secret during boot, critical parts have to be exposed

    Theoretically one could also put logs and cache on encrypted volumes. Maybe that could be some solution. I have in the past had /var/logs on separate partition, so it didn’t make /run out of space. Linux had no issue with that. But that still leaves kernel and OS exposed

    fully encrypted drive and this chain’s existence makes it easier to know that no one has tampered with my system

    The comment I responded to mentioned:

    needing to enter a password before the OS boots is a decision that makes Linux kind of awkward to use disk encryption with

    I don’t think you are talking about the same setup and vectors. Their point was to not have fully encrypted drive, so it boots without a prompt

  • OS encryption isn’t that important (verification is)

    I don’t think I’ve heard about some boot-time checksum verification of root partition. Doesn’t mean it does not exist, just that I can’t help here

    protecting user data

    My point is: if OS is not encrypted, it can be modified. And that verification idea, if is not stored under some encryption, could also be changed. Which means that by the time you put in your password to decrypt your home, you might be already running system that will nullify the protection. Encrypting only your data will only protect you in scenarios when someone snatches your device turned off

  • If you want ᜴ᚣᜩᗕ to get resolved to jake then bash will be a pain to use. I would use python

    For each ASCII letter create a list of non-ASCII characters that look similar. Then, for each word you want to match construct a regex

    1. dictionary = { ...'j': ['j', 'J', '𝚥', '𞋕', ...
    regex=''  
    for letter in 'jake':  
    regex += f'[{"".join(dictionary[letter])}]'  
    
    1. so after whole loop the regex would look a bit like (I’m cutting out a lot of characters to save on copy-pasting) [jJ𝚥𞋕][aA4@][kKᏦ🅺][eE3€]
    >>> import re  
    >>> r='[jJ𝚥𞋕][aA4@][kKᏦ🅺][eE3€]'  
    >>> re.fullmatch(r, 'jake')  
    <re.Match object; span=(0, 4), match='jake'>  
    >>> re.fullmatch(r, 'joke')  
    >>>  
    

    In general the group of problems that you are touching here is https://en.wikipedia.org/wiki/String_metric but I’m not sure if there is an algorithm that would be so “visual” matching