• Deceptichum@quokk.au
    link
    fedilink
    English
    arrow-up
    1
    ·
    5 months ago

    Damn, guess if you want reddit data to train your AI that you’ll need to pay Spez for access.

    • tal@lemmy.today
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      5 months ago

      It’s important for people writing papers and such who need to cite material.

      I wonder if there’s some way to use the TLS certificate to get a cryptographically-signed copy of a webpage with timestamp that someone could later validate as having been downloaded on that date. I don’t know if existing TLS libraries are capable of that. Like, Web browser menu option “Store cryptographically-signed webpage”. Absent a later certificate compromise, I’d think that that’d at least provide people a way to credibly say “this is really what was on that webpage on August 15th, 2026”. Like, you’d have to save a copy of the TLS session and then have libraries that could read and validate an already-generated session. The timestamp is already embedded in the session.

      Some protocols, like OTR, are designed to specifically not allow that, but AFAIK, TLS could.

      EDIT: Well, technically the timestamp is gonna be during the handshake, not tied to the HTTP request internal to the TLS session. It might be possible to game that by establishing a TLS session, holding it open without activity, and issuing a request much later. I’d think that that’d potentially be disallowed by Web servers one way or another, since otherwise you could probably do a denial-of-service attack by holding open a lot of sessions for a long time.

      EDIT2: Oh, wait, no, shouldn’t be an issue, because the HTTP Date response header is gonna have a timestamp tied to the response.