• 2 Posts
  • 22 Comments
Joined 11 days ago
cake
Cake day: December 7th, 2024

help-circle



  • Also, this is a good tip in general:

    You can’t fetch title and thumbnail for YouTube links that starts with youtu.be. This applies not only in Lemmy, but other websites too, for example Twitter (even before Elon’s era).

    This is, IMO due to the link doesn’t really contain HTML and OpenGraph metadata. It only redirects you. If you try cURL-ing the youtu.be link, you will see that it doesn’t have a response body, only some header including Location header that will tell you where to redirect. The response code itself is 301.


  • So I take a peek at the Lemmy’s source code.

    When you see a thumbnail and title of a URL (including YouTube video links) on Lemmy, what happens is that Lemmy server fetched the website HTML file and get the og:title and og:image (this convention is called OpenGraph protocol [1]) located inside <head> HTML tag.

    For example, for this video https://www.youtube.com/watch?v=MbYuEEU5e50, the OpenGraph metadata inside the HTML looks like this:

    <meta property="og:title" content="Flume - Lose It feat. Vic Mensa">
    <meta property="og:image" content="https://i.ytimg.com/vi/MbYuEEU5e50/maxresdefault.jpg">
    

    Try opening the “Create Post” page, then enter URL of a YouTube video. You might notice that a loading spinner will appear. During this, the frontend of Lemmy is actually creating a request to /api/v3/post/site_metadata?url=[insert your URL]. When the backend side of Lemmy receives the request from front end, it will run this code [2] to fetch the HTML and parse it to find the OpenGraph metadata. If the metadata is found, the backend will return it to front end, and you willl see recommended title and the “Thumbnail URL” field should be filled.

    In theory, this should work fine, regardless of any website. I personally tried to view the HTML code of a YouTube video and the HTML file indeed contains OpenGraph metadata.

    But it doesn’t for YouTube videos in Lemmy somehow. And I’m not sure.

    I’m suspecting YouTube thought that Lemmy servers are bots and because of that instead of receiving the HTML file of a YouTube video, instead it received a HTML file of a captcha page. I think I’m gonna try contacting admins of my instance to see whether YouTube URL works or not.


    1. https://ogp.me/ ↩︎

    2. https://github.com/LemmyNet/lemmy/blob/a2a5cb091a2892793179760275cb56529891bcb6/crates/api_common/src/request.rs#L203 ↩︎




  • Let me give you my opinion, specifically as a React developer, if you don’t mind.

    And let’s be clear: I intend this to be a constructive criticism. I hope you understand and don’t take it the wrong way.

    To be honest, I don’t know how good or bad federating one-way is. This is more of a “people” problem rather than a technical problem.

    But, to be honest, what I am bothered by, is the fact that the website doesn’t give an attribution in the UI about which instance certain users are from and which instance certain certain community are from.

    Take a look at this post: https://clubsall.com/posts/theyre-trying-to-charge-luigi-with-terrorism-imagine-that-qfF82

    The UI says that the post was posted by u[slash]BytesOnBikes. If I didn’t know better, I’d have assumed this was from a user from clubsall. But if you click the username, you realize that the link says u[slash]BytesOnBikes[at]slrpnk.net. I think this would be confusing as a user. What if there is the same user under the username BytesOnBikes from clubsall? At least if you include the instance name, user would know right away that both users are different. But if you didn’t include the instance name, I feel like this can be abused to impersonate user. This is a bad thing to happen to your website, don’t you agree?

    Now that we both understand that lack of attribution is a bad thing to clubsall… What’s stopping you from adding an instance name to the username? I’m sure the app has a way to know which instance certain users are from. From what I gather, I feel like this is as easy as appending a string in the code.

    I haven’t even talked about the community name on the UI. Or the ethicality of misleading attribution.









  • hono4kami@slrpnk.nettoFediverse@lemmy.worldDon't reply "just Google it"
    link
    fedilink
    English
    arrow-up
    226
    arrow-down
    5
    ·
    edit-2
    5 days ago

    As a software engineer…

    Don’t just say “just Google it”. Guide them to the documentation. Ask them about the detail of the question. If it’s an bug, try asking them if they can reproduce the bug.


    This reminded me of the time I’m looking for how to do certain things in a software. I found a reddit post asking about the same issue and this is the reply OP got:

    screenshot of a reddit thread where OP got a rude reply to search it themselves

    Here is the link: https://old.reddit.com/r/i3wm/comments/mupjsf/how_to_showhide_i3status_bar_taskbar/

    Imagine. You search the issue you have. Found the ONLY reddit thread that talks about this, and the ONLY thread that talks about the issue have NO USEFUL ANSWER and, worse, the only reply is TELLING YOU TO SEARCH IT YOURSELF. This got upvoted too 😭😭😭.

    Luckily, I found the solution (tbh the solution was there in the docs, but the wording wasn’t clear and it makes it hard to search) and I end up replying the OP the actual answer.

    So, this is a PSA for the fediverse: be nice. It’s free.

    While we’re still young, we have a chance to become a better forum.

    Also possibly an unpopular opinion: you shouldn’t downvote a question, even if it was asked multiple times. Guide them to the answer instead