3 ms·
Hate to be a hater too but this is pretty spot on. I too have been a Linux user since forever... 1993 to be precise. Linux is getting worse in certain respects
by api 10y ago
Hate to be a hater too but this is pretty spot on. I too have been a Linux user since forever... 1993 to be precise.
Linux is getting worse in certain respects on the server too, mostly due to the complexity explosion of systemd and its ilk and a complete lack of attempts to actually fix problems rather than piling on work-around after work-around after work-around.
SysV init was a nasty pile of arcane cruft, and to replace it the systemd crew somehow managed to engineer something more obtuse and arcane. That's an accomplishment I suppose. The whole thing feels "enterprise" from start to finish: awful counter-intuitive UX, monolithic, over-engineered, and I bet it completely destroys any hope of creating a Linux system hardened against local privilege escalation attacks.
Docker is brilliant satire in software form. It's fantastic and very useful, but for reasons that all relate to fundamental shortcomings and impedance mismatches in the core OS. Someone literally threw their hands up and said "F it lets just tar up Linux distributions and treat them like giant statically linked binaries!" and someone else said "wait... that's actually not a bad idea!" and here we are. It's a prank taken to production at scale, and the fact that it is so useful is a deep indictment of the whole ecosystem. If your roof leaks a tarp is useful too.
I've said it many times and I'll say it again: it is harder to make something simple than to make something complex. It require more effort, more knowledge, and more intelligence. Arcane or complex designs are a sign of inferior engineering.
Finally, there are far too many distributions. It makes shipping software for Linux a nightmare. The only reason anyone should ever create a distribution is to customize things for a specialized use case (e.g. embedded) or to explore some fundamentally novel approach to solving the problems outlined above (e.g. the Guix package manager, GoboLinux). Creating another distribution that is fundamentally just like all the others with the same essential approach to packaging and management and the like is needless fragmentation that only multiplies the effort required for every developer to target the platform.
On the 'meta' level: I think it's clear at this point that some of the classic assumptions of the FOSS community and especially the Linux wing of that community are wrong. Among these are Eric Raymond's whole thesis in "The Cathedral and the Bazaar." The reality seems to be more like "The Cathedral and the Mess."
... and JWZ is still right: https://www.jwz.org/doc/cadt.html https://www.jwz.org/doc/cadt.html
- anonbanker 10y agolinux user since 1998. systemd is a cancer. However, you might like OpenRC or runit (Calculate Linux or Void Linux are your friends). As for distributions, we as seasoned linux users have to take the initiative to adopt and popularize new distributions, the way we adopted Ubuntu en masse in 2006 out of nowhere. GalliumOS for chromebooks, Calculate Linux for laptops/desktops, and Void Linux (or gentoo hardened) for servers. When newbies ask for advice, we are doing them a disservice by suggesting Ubuntu or debian-based derivatives, despite the availability of binaries for everything, for the same reason that people suggesting RPM-based derivatives (Red Hat, CentOS, Fedora) were doing a disservice to newbies. My 13yo daughter can get through a Calculate Linux, Manjaro, or Antergos install (though she needs help replacing systemd with OpenRC on the last two). Let's start promoting the new good distributions, now that the old ones have jumped the shark.
- gergles 10y agoYou may want to edit that to not link to jwz, since he referer detects and redirects HN to an NSFW image.
- slgeorge 10y agoThe issue with the "Cathedral and the Bazaar" is that he's right for an individual piece of software, but wrong for an integrated system. Can you get the best editor, best grep, best <insert technical tool> with FOSS? The answer is yes. Multiple implementations, and finally the best one "wins", at least for a while. But you can you get the best system? No, because you land up with multiple incompatible implementations that don't quite fit together. OSX is the most similar option, and it wins because of consistency - consistency is forced on all developers by an overall authority. Keyboard shortcuts are a classic example, OSX has one set, on Linux you get different ones depending on the toolkit used GNOME, KDE, Xwin and a lot of other defunct ones! In the Linux world the desktop environments tried to define their own standards (GNOME and KDE in particular) and it didn't work because users run applications from other tool kits; then they tried through freedesktop.org but it just didn't get/have the buy-in. You've also always had the problem of the major distributions competing (wanting to differentiate) with each other and causing friction. And finally, these days there's just not that much energy in the desktop paradigm on Linux: there are very few "application" focused desktop developers now. Agreed, that the creation of distributions is also a problem. It seems to be a test that every 'power administrator' does, a bit like every developer creates a text editor/irc client. Somewhere along the line Linux user/developers got the idea that creating their own distribution was cool. Distributions are easy to do at the beginning, and then increasingly hard. It's a separate issue that you don't get the "best <insert consumer app>" because the people that develop open source don't understand 'consumer apps'.