RedWizard [he/him]

  • 0 Posts
  • 9 Comments
Joined 3 years ago
Cake day: June 9th, 2023
  • RedWizard [he/him]@lemmy.mltoMemes@lemmy.ml•-1 Fedi Social Credit
    391·
    2 days ago

    Ok, you have this completely off base about it “not doing anything”. While I might be wrong about it defederating, what this code ACTUALLY does is rate other instances defederation lists based on this hard coded list. Let me explain:

    The site_instance_chooser_view() function in /app/api/alpha/views.py provides a JSON representation of the current site’s metadata for the instance chooser feature. This feature allows users to browse and compare different Fediverse instances before choosing one to join.

    Within site_instance_chooser_view(), the defed_list variable is defined as follows (lines 1148‑1151):

    defed_list = BannedInstances.query.filter(or_(BannedInstances.domain == 'hexbear.net',
                                              BannedInstances.domain == 'lemmygrad.ml',
                                              BannedInstances.domain == 'hilariouschaos.com',
                                              BannedInstances.domain == 'lemmy.ml')).order_by(BannedInstances.domain).all()

    This query retrieves only four specific domains from the BannedInstances table:

    1. hexbear.net
    2. lemmygrad.ml
    3. hilariouschaos.com
    4. lemmy.ml

    The resulting list is used to populate the defederation field in the returned JSON (line 1187):

    'defederation': list(set([instance.domain for instance in defed_list])),

    The defederation field is part of the site metadata returned by the API endpoint /api/alpha/site/instance_chooser. This endpoint is called by the instance‑chooser UI (/auth/instance_chooser) when a user clicks “More” on an instance card.

    The template app/templates/auth/instance_chooser.html uses the defederation list to compute a defederation quality rating. The rating is based on how many of the four watched domains are blocked:

    • ≥3 blocked → “Good”
    • 2 blocked → “Ok”
    • 1 blocked → “Minimal”
    • <1 blocked → “Negligent”

    This rating is displayed in the instance details modal under the “Defederation” label (line 114 of the template).

    The UI also contains commented‑out code (lines 124‑130) that would show individual status indicators for each of the four domains, but this is currently disabled.

    This is problematic for a number of reasons, most of all is that this rating that it generates is NOT transparent to the user. This page is used on PieFeds main page when you go to register, it’s part of the instance picker. The defederation rating under More is where this shows up. For instance, this means that instances like anarchist.nexus have a “OK” rating but instances like multiverse.soulism.net have a “GOOD” rating.

    Anarchist.nexus has an “OK” raiding because they block Lemmygrad.ml (socialist) and hilariouschaos.com (MAGA instance)

    multiverse.soulism.net has a “GOOD” rating because they block Lemmygrad.ml (socialist), Hexbear.net (socialist), lemmy.ml (operated by open communists).

    So the Defederation rating has an OBVIOUS BIAS that isn’t explained to the users at all. Not only is the bias not explained it doesn’t even contain all of the FASCIST INSTANCES IN ITS CALCULATION.

  • RedWizard [he/him]@lemmy.mltoMemes@lemmy.ml•-1 Fedi Social Credit
    39·
    2 days ago

    In the admin back end video produced to show the features of the software, these sites were said to be “defaults” by the software creator, and they are prefilled in. You can change them after the fact, this is true, but if you simply spin up the instance and never touch those settings they are defederated. I know this is true, because I am in contact with an admin who manages a PieFed instance that is federated with Hexbear, they had to remove the Hexbear defederation after initial setup.

  • RedWizard [he/him]@lemmy.mltoMemes@lemmy.ml•-1 Fedi Social Credit
    34·
    2 days ago

    This is an incredible thread. This is the kind of ‘fast-and-loose’ nonsense that really lives at the heart of PieFed. Most people’s complaints about Lemmy have nothing to do with the software. The software does exactly what it says it does, and it does it in a very stable way. PieFed however, will suddenly add code to remove the Unicode Þ from comments because the repo maintainer things a user is being annoying. Why do would anyone want to hitch their wagon to such a thing?

  • RedWizard [he/him]@lemmy.mltoMemes@lemmy.ml•-1 Fedi Social Credit
    931·
    2 days ago

    There are all kinds of fun stuff in the Piefed code. Allow me to dredge up a comment I made recently:

    @edie@lemmy.encryptionin.space was looking at PieFed code the other week, and I ended up taking a look at it too. Its great fun to sneak a peak at.

    For example, you cannot cast a vote on PieFed if you’ve made 0 replies, 0 posts, AND your username is 8 characters long:

        def cannot_vote(self):
        if self.is_local():
            return False
        return self.post_count == 0 and self.post_reply_count == 0 and len(
            self.user_name) == 8  # most vote manipulation bots have 8 character user names and never post any content

    If a reply is created, from anywhere, that only contains the word “this”, the comment is dropped (CW: ableism in the function name):

    def reply_is_stupid(body) -> bool:
    lower_body = body.lower().strip()
    if lower_body == 'this' or lower_body == 'this.' or lower_body == 'this!':
        return True
    return False

    Every user (remote or local) has an “attitude” which is calculated as follows: (upvotes cast - downvotes cast) / (upvotes + downvotes). If your “attitude” is < 0.0 you can’t downvote.

    Every account has a Social Credit Score, aka your Reputation. If your account has less than 100 reputation and is newly created, you are not considered “trustworthy” and there are limitations placed on what your account can do. Your reputation is calculated as upvotes earned - downvotes earned aka Reddit Karma. If your reputation is at -10 you also cannot downvote, and you can’t create new DMs. It also flags your account automatically if your reputation is to low:

    PieFed boasts that it has “4chan image detection”. Let’s see how that works in practice:

                if site.enable_chan_image_filter:
                # Do not allow fascist meme content
                try:
                    if '.avif' in uploaded_file.filename:
                        import pillow_avif  # NOQA
                    image_text = pytesseract.image_to_string(Image.open(BytesIO(uploaded_file.read())).convert('L'))
                except FileNotFoundError:
                    image_text = ''
                except UnidentifiedImageError:
                    image_text = ''

                if 'Anonymous' in image_text and (
                        'No.' in image_text or ' N0' in image_text):  # chan posts usually contain the text 'Anonymous' and ' No.12345'
                    self.image_file.errors.append(
                        "This image is an invalid file type.")  # deliberately misleading error message
                    current_user.reputation -= 1
                    db.session.commit()
                    return False

    Yup. If your image contains the word Anonymous, and contains the text No. or N0 it will reject the image with a fake error message. Not only does it give you a fake error, but it also will dock your Social Credit Score. Take note of the current_user.reputation -= 1

    PieFed also boasts that it has AI generated text detection. Let’s see how that also works in practice:

    # LLM Detection
        if reply.body and '—' in reply.body and user.created_very_recently():
            # usage of em-dash is highly suspect.
            from app.utils import notify_admin
            # notify admin

    This is the default detection, apparently you can use an API endpoint for that detection as well apparently, but it’s not documented anywhere but within the code.

    Do you want to leave a comment that is just a funny gif? No you don’t. Not on PieFed, that will get your comment dropped and lower your Social Credit Score!

            if reply_is_just_link_to_gif_reaction(reply.body) and site.enable_gif_reply_rep_decrease:
            user.reputation -= 1
            raise PostReplyValidationError(_('Gif comment ignored'))

    How does it know its just a gif though?

    def reply_is_just_link_to_gif_reaction(body) -> bool:
    tmp_body = body.strip()
    if tmp_body.startswith('https://media.tenor.com/') or \
            tmp_body.startswith('https://media1.tenor.com/') or \
            tmp_body.startswith('https://media2.tenor.com/') or \
            tmp_body.startswith('https://media3.tenor.com/') or \
            tmp_body.startswith('https://i.giphy.com/') or \
            tmp_body.startswith('https://i.imgflip.com/') or \
            tmp_body.startswith('https://media1.giphy.com/') or \
            tmp_body.startswith('https://media2.giphy.com/') or \
            tmp_body.startswith('https://media3.giphy.com/') or \
            tmp_body.startswith('https://media4.giphy.com/'):
        return True
    else:
        return False

    I’m not even sure someone would actually drop a link like this directly into a comment. It’s not even taking into consideration whether those URLs are part of a markdown image tag.

    As Edie mentioned, if someone has a user blocked, and that user replies to someone, their comment is dropped:

    if parent_comment.author.has_blocked_user(user.id) or parent_comment.author.has_blocked_instance(user.instance_id):
    log_incoming_ap(id, APLOG_CREATE, APLOG_FAILURE, saved_json, 'Parent comment author blocked replier')
    return None

    For Example:

    (see Edies original comment here)

    More from Edie:

    Also add if the poster has blocked you! It is exactly as nonsense as you think.

    Example:

    I made a post in testing@piefed.social from my account testingpiefed@piefed.social, replied to it from my other testingpiefed@piefed.zip account. Since the .social account has blocked the .zip, it doesn’t show up on .social, nor on e.g. piefed.europe.pub.

    I then made a comment from my lemmy.ml account, and replied to it from my piefed.zip account, and neither .social, nor europe.pub can see my .zip reply, but can see my lemmy.ml comment!

    [ Let me add more clarity here: what this feature does is two things. On a local instance, if you block someone who is on your instance, they cannot reply to you. However, this condition is not federated (yet, it would seem), and so, to get around this “issue”, the system will drop comments from being stored in the PieFed database IF the blocked user is remote. This means you end up with “ghost comment chains” on remote instances. There is NEW code as of a few weeks ago, that will send an AUTOMATED mod action against blocked remote users to remove the comment. So long as the community is a local PieFed community, it will federate that mod action to the remote server, removing the comment automatically. For PieFed servers, eventually, they would rather federate the users block list (that’s fair), but it would seem this code to send automated mod actions to remove comments due to user blocks is going to stay just for the Lemmy Piefed interaction. I don’t really understand why the system simply doesn’t prevent the rendering of the comment, instead of stopping it from being stored. It knows the user is blocked, it already checks it, it should then just stop rendering the chain of comments for the given user, prevent notifications from those users, etc. ]

    But wait! There’s More!

    All this to say. Piefed is a silly place, and no one should bother using its software.