• deborah@awful.systems
    link
    fedilink
    English
    arrow-up
    14
    ·
    edit-2
    17 days ago

    when you do not yet have (1) customers, (B) unit tests, (ג) developers who can write their own code, or (IV) exception handling, the term-of-art that comes to mind for doing anything besides auto-incrementing primary keys is YAGNI. (Especially because nobody who is making thoughtful, careful database tuning decisions is using chat-gippity to convert their models. And more to the point, they aren’t using SQLAlchemy of all things to make large, distributed applications that need UUID primary keys.)

    • Hexarei@programming.dev
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      3
      ·
      edit-2
      17 days ago

      Oh for sure, the article folks are inept and absolutely not the people I was talking about. I’m just talking about stuff more like Discord or Steam that are huge distributed systems that don’t use centralized databases.

      Edit: that don’t use centralized databases. I blame the ADHD.

      Edit 2: I am agreeing with this person

      • ebu@awful.systems
        link
        fedilink
        English
        arrow-up
        9
        ·
        17 days ago

        I’m just talking about stuff more like Discord or Steam that are huge distributed systems that don’t use databases.

        huh???

        • Hexarei@programming.dev
          link
          fedilink
          English
          arrow-up
          8
          ·
          17 days ago

          Whoops, I flubbed that message hard and didn’t catch it at the time: Meant to say “don’t use centralized databases.” They definitely use databases lmao. No idea how I screwed that message up so hard. I blame ADHD for not proofreading.

          Just so we’re on the same page, let me be more specific. I’m saying the individuals in the article were making terrible decisions. Lots of them.

          I am also saying that UUIDs are good primary keys for very specific purposes: Large, distributed systems that handle large amounts of small data, powered by databases like Cassandra that are designed to handle millions of record insertions per hour across several hundred nodes, to the point where inserts are very likely to happen at the exact same time on two different replicas of the same schema.

          Hope that makes more sense than my previous flub. lol

          • ebu@awful.systems
            link
            fedilink
            English
            arrow-up
            5
            ·
            edit-2
            17 days ago

            okay that’s a little more sensible lol

            i think the original comment that this thread is in reply to is avoiding non-monotonic UUIDs. i don’t think anyone is contesting that autoincrementing ints create headaches when trying to distribute the database

        • froztbyte@awful.systems
          link
          fedilink
          English
          arrow-up
          5
          ·
          17 days ago

          See, reason being is they use aethernet - that’s the only way you get to get scale it like this. Without that, communication and storage would just be impossible!