↵return

trainwreck design

another pufferfish drawn by rekka

trainwreck design 2024-08-14 the year is 2024, and i want to know my disk space usage. i recall a simple command i've used probably tens of thousands of times. df -h the year is 2024. instead of sensible output, i get... this: Filesystem Size Used Avail Use% Mounted on dev 7.7G 0 7.7G 0% /dev run 7.8G 1.5M 7.8G 1% /run efivarfs 184K 150K 30K 84% /sys/firmware/efi/efivars /dev/nvme0n1p2 234G 193G 30G 87% / tmpfs 7.8G 860K 7.8G 1% /dev/shm tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-udev-load-credentials.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-journald.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup-dev-early.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-sysctl.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup-dev.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-vconsole-setup.service tmpfs 7.8G 20M 7.7G 1% /tmp /dev/nvme0n1p1 511M 296M 216M 58% /boot tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup.service tmpfs 1.6G 32K 1.6G 1% /run/user/1000 tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-resolved.service my eyes glaze over. somehow, i can never seem to find the disk i actually care about. somehow, "mount" is even worse: proc on /proc type proc (rw,nosuid,nodev,noexec,relatime) sys on /sys type sysfs (rw,nosuid,nodev,noexec,relatime) dev on /dev type devtmpfs (rw,nosuid,relatime,size=8063880k,nr_inodes=2015970,mode=755,inode64) run on /run type tmpfs (rw,nosuid,nodev,relatime,mode=755,inode64) efivarfs on /sys/firmware/efi/efivars type efivarfs (rw,nosuid,nodev,noexec,relatime) /dev/nvme0n1p2 on / type ext4 (rw,relatime) securityfs on /sys/kernel/security type securityfs (rw,nosuid,nodev,noexec,relatime) tmpfs on /dev/shm type tmpfs (rw,nosuid,nodev,inode64) devpts on /dev/pts type devpts (rw,nosuid,noexec,relatime,gid=5,mode=620,ptmxmode=000) cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,nsdelegate,memory_recursiveprot) pstore on /sys/fs/pstore type pstore (rw,nosuid,nodev,noexec,relatime) bpf on /sys/fs/bpf type bpf (rw,nosuid,nodev,noexec,relatime,mode=700) systemd-1 on /proc/sys/fs/binfmt_misc type autofs (rw,relatime,fd=39,pgrp=1,timeout=0,minproto=5,maxproto=5,direct,pipe_ino=1374) mqueue on /dev/mqueue type mqueue (rw,nosuid,nodev,noexec,relatime) hugetlbfs on /dev/hugepages type hugetlbfs (rw,nosuid,nodev,relatime,pagesize=2M) debugfs on /sys/kernel/debug type debugfs (rw,nosuid,nodev,noexec,relatime) tracefs on /sys/kernel/tracing type tracefs (rw,nosuid,nodev,noexec,relatime) tmpfs on /run/credentials/systemd-udev-load-credentials.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) tmpfs on /run/credentials/systemd-journald.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) fusectl on /sys/fs/fuse/connections type fusectl (rw,nosuid,nodev,noexec,relatime) tmpfs on /run/credentials/systemd-tmpfiles-setup-dev-early.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) configfs on /sys/kernel/config type configfs (rw,nosuid,nodev,noexec,relatime) tmpfs on /run/credentials/systemd-sysctl.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) tmpfs on /run/credentials/systemd-tmpfiles-setup-dev.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) tmpfs on /run/credentials/systemd-vconsole-setup.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) tmpfs on /tmp type tmpfs (rw,noatime,inode64) /dev/nvme0n1p1 on /boot type vfat (rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,utf8,errors=remount-ro) tmpfs on /run/credentials/systemd-tmpfiles-setup.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) binfmt_misc on /proc/sys/fs/binfmt_misc type binfmt_misc (rw,nosuid,nodev,noexec,relatime) tmpfs on /run/user/1000 type tmpfs (rw,nosuid,nodev,relatime,size=1615816k,nr_inodes=403954,mode=700,uid=1000,gid=1000,inode64) portal on /run/user/1000/doc type fuse.portal (rw,nosuid,nodev,relatime,user_id=1000,group_id=1000) tmpfs on /run/credentials/systemd-resolved.service type tmpfs (ro,nosuid,nodev,noexec,relatime,nosymfollow,size=1024k,nr_inodes=1024,mode=700,inode64,noswap) jesus christ. and if you think that the above output is ugly, let me assure you, it's even worse in a terminal. each 150+ character line wraps around my buffer like a boa constrictor, completely suffocating any sense of meaning. ,=e `-. _,-' my journey to know how much disk space i had left has left me confused. annoyed. i feel disrespected in my own ~home. and i don't like being disrespected. especially by command line applications. after all, this is MY terminal. MY userspace. a program just waltzes in and spews shitty output all over MY tty? HELL NO oh, but tmpfs's are so very useful? fuck you. oh, efivars are important becau- FUCK YOU. I AM A REGULAR LINUX USER JUST SHOW ME MY DISKS PWEASE THIS WAS SO MUCH BETTER 10 YEARS AGO I AM SO SICK OF "TRAINWRECK DESIGN" O DEAR LORD, IS THERE ANY END TO THIS TORMENT? OO O o o o... _______________ _________________ O ____ | | | | ][_n_i_| ( ooo___ | trainwreck | | design | (__________|_[______]_|_____________|_|_______________| 0--0--0 0 0 0 0 0 0 as open ecosystems grow, they tend to experience what i like to call "trainwreck design" basically: when many programs are independently designed and developed & then smashed together, user experience suffers. things feel wrong because users can sense the lack of coherent vision. tiny papercuts. take the GNU/Linux ecosystem: "trainwreck design" happens there all of the time. programs are designed independently & smashed together, everyone spritzes holy water on their keyboards, adds 2,000 more lines of lua to their emacs config, and things mostly sort of keep stumbling forward. but this results in programs like df. or mount. they grow weird appendages, have millions of esoteric flags, and worst of all, they just FEEL obviously wrong. "trainwreck design" is not intentional design. people have tried using consensus-based design, but consensus doesn't really work in the design realm. BAZAAR DESIGN THEORY when an open, bazaar-like[0] community grows too large, it becomes impossible to insist on a coherent design philosophy, because getting that many people to agree on vision leads to slowdowns, disagreements, forks, etc. in other words, bazaar-like communities result in bazaar-like design. ____ | | .----------------------------------------------.| | | | | | A camel is a horse designed by a committee | | | ._|____|_. '---------------------------------------------\\|o_o | |:_/ | // \ \ (| [4] | ) i actually think the above is completely false. i have my own theory: bazaar-style development does not necessitate committee-style design. a bazaar can in fact have a unified vision. but it must be intentional from the beginning. this is already happening - many projects that are very much bazaar-like in nature have a strong visionary + designer (or several), who typically hold leadership positions. but it's rare, and the territory is pretty new. historically, this position has affectionately been called the~ ██████ ██████ ███████ ██ ██ ██ ██ ██ ██ ██ ██████ ██ ██ █████ ██ ██ ██ ██ ██ ██ ██ ██████ ██████ ██ ███████ "benevolent dictator for life" for a good example, see the hare[1] project's BDFL page. Drew Devault runs the hare project, and wrote that page. my conclusions: - Drew is inclusive of the broader hare community - Drew has the final say over certain decisions - Drew will not "abuse" his position - Drew has "the vision" - ... - Drew is the designer. try reading hare's style[2] guide and this[3] post on hare's design philosophy. hare is clearly a designed project. "Hare Is Boring" is an assertion made by someone with a strong innate sense of design. a new contributor can immediately read, digest, and align with Hare's vision. they can contribute code in an expected style with very little oversight. they can get excited about hare, because when design is this up-front, it's compelling. exciting. design requires vision. direction. involvement. design does _not_ require the designer to do all of the work, or be a gate through which all work flows. the designer does _not_ have to be a dictator. this sort of setup is neither a bazaar nor a cathedral - i believe it's somewhere in between. the cathedral, the bazaar, and... ████████╗██╗ ██╗███████╗ ██╗ ██╗██╗██╗ ██╗███████╗ ╚══██╔══╝██║ ██║██╔════╝ ██║ ██║██║██║ ██║██╔════╝ ██║ ███████║█████╗ ███████║██║██║ ██║█████╗ ██║ ██╔══██║██╔══╝ ██╔══██║██║╚██╗ ██╔╝██╔══╝ ██║ ██║ ██║███████╗ ██║ ██║██║ ╚████╔╝ ███████╗ ╚═╝ ╚═╝ ╚═╝╚══════╝ ╚═╝ ╚═╝╚═╝ ╚═══╝ ╚══════╝ .' '. __ . . . (__\_ . . . -{{_(|8) ' . . ' ' . . ' (__/ there is third style of software design that is often conflated with the other two. i'm calling it "the hive", and it's a bazaar/cathedral hybrid. hives have one or more leaders who influence direction and vision, but who do _not_ act as dictators. in other words, hive leaders inspire design through vision - not by telling people what to do. being a hive leader is not fun. they have stressful, emotional responsibilities: - produce vision and guidance for the project - help people understand the vision - organize people into roles - resolve interpersonal disputes - provide MORE guidance when people disagree - watch all of the trains - try not to burn the fuck out people involved with hive projects make decisions autonomously, but with shared understanding and vision. good hive leaders dramatically reduce the likelihood of trainwreck design, and they keep the important communal "vibe of the bazaar" alive. also, you get to call yourself the "queen bee" .' '. __ . . . (__\_ . . . -{{_(|8) ' . . ' ' . . ' (__/ tldr: bazaar design: bad bazaar community: good cathedral design: good cathedral community: bad hive design: good hive community: also good if somebody is watching all of the trains and showing other people how to watch then maybe some of the trains won't collide love, jes P.S. the year is 2035. you run "df -h" on your holo-deck. Filesystem Size Used Avail Use% Mounted on dev 7.7T 0 7.7T 0% /dev run 7.8T 1.5G 7.8T 1% /run efivarfs 184K 150K 30K 84% /sys/firmware/efi/efivars /dev/qua0n1p2s4g8z7@ 234T 193T 30T 87% / tmpfs 7.8G 860K 7.8G 1% /dev/shm tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-udev-load-credentials.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-journald.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-systema-systemb.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup-dev-early.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-sysctl.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-kitchen-sink.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup-dev.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-vconsole-setup.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-dishwasher.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-mothership.service tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-taxes.service tmpfs 7.8G 20M 7.7G 1% /tmp qtmpfs 7.8G 20M 7.7G 1% /tmp-quantum-rebind /dev/qua0n1p38fuqff@ 234T 193T 30T 87% / /dev/qua0n1p1984dja@ 511M 296M 216M 58% /boot tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-tmpfiles-setup.service tmpfs 1.6G 32K 1.6G 1% /run/user/1000 tmpfs 1.0M 0 1.0M 0% /run/credentials/systemd-resolved.service /dev/quantum0 dead 0 1.0T 0% /quantum/superposition/cat/0 /dev/quantum1 alive 0 1.0T 0% /quantum/superposition/cat/1 cybermem 1P 750T 250T 75% /run/cybermemory/cybermem-setup cybermem 1P 750T 250T 75% /run/cybermemory/cybermem-init cybermem 1P 750T 250T 75% /run/cybermemory/cybermem-run cybermem 1P 750T 250T 75% /run/cybermemory/cybermem-RUN virtualsocks 64G 2G 62G 3% /run/vsockies (\(\ (-.-) (")(")* [0]: https://en.wikipedia.org/wiki/The_Cathedral_and_the_Bazaar [1]: https://archive.is/H6qQA [2]: https://archive.is/ypKFZ [3]: https://archive.is/9762X [4]: ty to dylan araps for this ascii thingy thanks to Joan Stark for much of the bee ascii art

follow me on mastodon!

last updated 2024-08-14T00:00:00.000Z