1. 37
  1.  

  2. 9

    At this point I live in perpetual, mild trepidation that the people who maintain the Galapagos island of software I rely on will stop caring, stranding me in desktop Linux limbo.

    This is my primary gripe with Linux distros.

    At one point, in a vain attempt to inure myself from other people’s spotty support of my tendency as a hoarder of old junk, I tried experimenting with linuxfromscratch.org to estimate how deep I’d have to go, to truly roll my own system, with near total control over absolutely everything.

    I lost about four weekends putting together sketches of working builds with the intent of modelling a standard machine image for myself, before I realized the only way this might ever work, is if I turned it into a full time job.

    1. 9

      This is a core reason I like NixOS. Every install of NixOS includes the instructions on how to recreate it again, with full freedom and ability to dive in and make changes. There are no special tools or separate processes to repackage something.

      1. 10

        I lost about four weekends putting together sketches of working builds with the intent of modelling a standard machine image for myself, before I realized the only way this might ever work, is if I turned it into a full time job.

        and instead all you needed was this ;)

        1. 3

          Yeah I have had that experience with Linux From Scratch too.

          Probably my programmer’s bias, but I thought that writing code accounted for most of the work in open source. No the packaging is also a lifetime of work too! We take for granted an incredible amount of work.

          Although I honestly wonder if a reliable and fast distributed build system, and a consistent test environment for packages would help.

          1. 3

            I doubt it. The cost of release engineering is mostly one of finding and supporting competent people who are willing to do it instead of something perhaps perceived to be a bit more exciting.

            1. 1

              Well, I would say it would be more exciting if you weren’t waiting around for builds all the time. And also if the results were actually verified with automated tests… there are a ton of buggy Debian packages with no tests.

              For all the confusion in the term, “devops” is a step in the right direction… using software engineering practices for what used to be perceived as manual sys admin work.

            2. 2

              A lot of people want to solve this for NixOS, and usually the biggest problem is trust:

              • trusting the builders are not malicious
              • trusting the packagers are not malicious
              1. 1

                Yeah that is a good point. I played with building something like this, and in retrospect I did end up spending a lot of time thinking about security. If you want to share computation, then you have to trust the people and computers doing it!

          2. 6

            “Scripts for tearing apart PDFs, adding my signature to a page, and stitching them back together again.”

            I want those!

            1. 3

              LD_PRELOAD and a very lightly patched Go implementation go a long way for testing clock skew locally

              1. 1

                That’s not useful if you are trying to test software written in an arbitrary language and possibly without access to source code.

                1. 8

                  That is exactly what LD_PRELOAD is useful for. People tend to write databases in things that call out to shared libraries. Some stuff is Go. It has been extremely useful for me as someone who fault injects databases, sometimes using lxc.

              2. 3

                i am somewhere between appalled that he’d go so long with a labelless keyboard and ruefully aware that i would probably gripe but do nothing about it too :)

                1. 2

                  It’s more amazing to me that you could go that long without developing muscle memory.

                  1. 5

                    I can touch-type, but I still look at the keyboard when I type complex passwords. I think my muscle memory is linked to sequences of key presses more than to specific characters. So I can type a sentence, but give me a random sequence of letters, numbers, and symbols and I’ll mess it up if I don’t glance at the keyboard.

                    1. 2

                      I have my keyboard arranged in alphabetical order, mostly to troll anybody else who sits down at my keyboard. Oddly enough, I still find it useful to look at the keyboard when typing a complex password, even though the letters on the keys have nothing to do with the character that they result in.

                2. 3

                  The motherboard is wonky and refuses to find half the disks on boot. You can crash the box by using certain USB ports. We have a complicated relationship.

                  I’m curious about this. You can buy a stable, reliable computer with 48 cores and 128GB of ECC RAM completely off the shelf — Dell/HP/etc sell rackmount or tower servers with these configurations.

                  Is using a really powerful but unreliable desktop computer a net productivity advantage relative to using a small reliable desktop + SSHing into a big reliable server? I appreciate that remote debugging is often not as nice as local debugging, but on the other hand remote debugging has some nice side benefits like the fact that the UI that you’re using doesn’t go unresponsive when the machine gets loaded.

                  I can totally sympathise if it turns out that the root cause of this was just that Kyle just really really wanted a really overpowered computer out of sheer nerdlust.

                  1. 8

                    sheer nerdlust

                    I know it’s a time-honored tradition to armchair-architect strangers’ technical decisions without regard for the privileges of context or experience, but fuck it, I’ll bite.

                    I used to rebuild and rack servers for a living, and considered that here, but ultimately decided I wanted a workstation.

                    It’s quieter; I didn’t feel like having a screaming banshee 2u sitting in my tiny SF bedroom. It means owning one computer instead of two, which is cheaper, takes up less space, and cuts down on my time spent doing stupid sysadmin stuff. It’s also way less of a pain in the ass to work with than the janky-ass combination of remote filesystems, SSH tunnels, rsync hacks, X forwarding, and Yourkit injection that I have to use with remote Jepsen clusters.

                    1. 6

                      Thank you for replying! I’m sorry if I came off as insulting. Edit: I apologise for insulting you. That was not my intention.

                      It’s quieter

                      I appreciate the noise issue. I’m used to 1U servers being awful for it because the fans are small so the noise is high-pitched, haven’t gotten my hands on 2U or up to see if they’re much quieter. I thought tower servers were supposed to be no worse than desktops in this regard? Since they’re not that different and can use similarly huge fans?

                      janky-ass combination of remote filesystems, SSH tunnels, rsync hacks, X forwarding, and Yourkit injection

                      Ouch. Good point, avoiding that mess is worth a lot of effort.

                    2. 1

                      My guess is that its some kind of whitebox. I’ve never had good lucky with them, and always some kind of jank. I replaced my whitebox server with an SFF business desktop and it’s been far better in stability.