1. 20
  1. 8

    Dear God, they were talking about GNU Hurd even back then…

    1. 3

      for 30 years they’ve been talking about it as if it’s almost ready.

      1. 3

        I always used to joke by substituting “when hell freezes over” with “when Duke Nukem Forever is ported to GNU Hurd”. I was a little bit sad when Duke Nukem Forever was finally released as it killed my joke :-(

        1. 1

          Imagine if we got +net fusion power before GNU Hurd?

          1. 1

            I don’t think this is quite fair, the GNU project have been mostly describing it as rather complete and usable since 2015. They describe it as an interesting development project, suitable for further development and technical curiosity, rather than necessarily a “production” OS, but the idea that HURD is perpetually nearly ready is fictional.

            1. 6

              Hurd has been making progress somewhat more slowly than the baseline requirements for a useful OS have been progressing. It passed the point of an early ‘90s *NIX kernel quite a long time ago (basic filesystem, able to run an X server, small number of drivers) and had quite a lot of nice features. The design of Hurd means that things like containers are supported automatically (anything that’s a global namespace in a traditional *NIX is just a handle that you get from the parent, so creating an isolated namespace is trivial. I still find it an interesting example of worse-is-better that the overwhelming majority of container deployments are on the one contemporary system that doesn’t have native support for containers.

              1. 1

                For me the biggest problem with contemporary Hurd is also the one I’m unqualified to fix: the drivers are all Linux 2.2-2.6 era. Given a more modern filesystem and newer drivers it’d be quite liveable.

          2. 2

            I thought a nod in the announcement for Linux, but no. In very early email threads, Linus wrote things like “this might be a fun toy to play with, until Hurd is usable, in a year or two.”

          3. 7

            many companies that are household names today–Netscape, Yahoo, Excite–simply did not exist.

            It’s funny how these things go. I didn’t realise this was from 1999 until this point.

            1. 5

              The original debate is from 1992, back then the Internet still had an Acceptable Use Policy that prohibited commercial activity.

            2. 3

              Do note that this debate continued, and Linus Torvalds is still wrong on this.

              1. 2

                Not disagreeing, but that post dates to 2006.

                1. 1

                  As far as I am aware, Linus is yet to acknowledge the superiority of the microkernel approach.

                  1. 8

                    While I find microkernels interesting (albeit only one facet of the discussion in kernel design, it overshadows others), I’m not sure if you’re aware how smug/condescending you sound in this comment, and others I’ve seen from you on this topic. It suffocates the potential for discussion/debate and sets a bad atmosphere for it.

                    1. 3

                      Bad atmosphere or not, I think everybody who has ever had a kernel problem in production shares the sentiment.

                      What I would give for the ability to gcore and restart a misbehaving process, instead of the entire machine with several dozen tenants on it. It makes development and security updates easier too.