• fishpen0@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    arrow-down
    1
    ·
    9 hours ago

    Right up until you are doing compliance and governance and you realize docs are actually a terrible terrible source of truth for any automated systems. We’re 3 years into a project at a healthcare company to rip google sheets and docs out of our apps and replacing them with Postgres, bigquery, dbt and dagster.

    It’s simply not okay to have your database be something anyone with write access to a doc can fuck up a formula by accident on. Your medical bills being maintained by random formulas on dozens of linked spreadsheets maintained by hand by random people on different teams is part of why they are impossible to unwind. By the time someone audits it, it’s printing different numbers than when your bill was rendered and it’s version control doesn’t work to roll it back without breaking dozens of other things.

    • werefreeatlast@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 hours ago

      I’m in the engineering business. We have a PDM system that we check-in copies of component 3D models, PDF drawings and DOCs. Once your team has collaborated enough, you have a copy…once a week/day/hour depending on your preference. That way you can collaborate and keep frozen records and rev controlled documents.

      • fishpen0@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 hour ago

        Right. But you can’t do that in a live system like google docs. You can have a workflow to export copies, but the live doc is the one bigquery and linked docs utilize to function against your app. It’s actually a feature of the same tooling that makes using them like a database possible that causes it to not be versionable. So even if you export copies as you update it, you can’t move the system back to those copies without breaking other parts of the system.

        Other systems for modeling data have better version control for running parallel versions of models if you need to recover how data had been constructed in an older state. It’s an incredibly bad idea to do this with Google docs at scale