|
the only person worse than a professional software developer is a professional linux kernel developer
|
# ¿ Aug 14, 2020 21:42 |
|
|
# ¿ May 13, 2024 00:30 |
|
i used linux for my primary os at work for 15 years. RH, Gentoo, Ubuntu, and many weird variants. any claims that “a normal user would open a github issue on the relevant component and get the problem fixed for everyone” is a statement made by a moron who has no business declaring how any technology should work for anyone ever. this is also exactly the kind of person who is generally in charge of major decision making for whatever their corner fiefdom of technology is in the linux component stack. thats why i dont run linux for my primary os anymore. those people suck and im tired of their poo poo.
|
# ¿ Nov 10, 2021 13:11 |
|
KozmoNaut posted:OpenSUSE has been very good to me in that regard, so far. yeah i think if i had to again id go with a suse flavour they just seem to kinda actually care about the customer experience and making a good product as opposed to mostly being pedantic assholes. i know some of the suse folks are in fact pedantic assholes, but they dont get to set the tone of the distro overall for some reason.
|
# ¿ Nov 10, 2021 13:54 |
|
infernal machines posted:this is the guy that managed to build a wholly unsupported nested raid with all the redundancy of raid 0, completely by accident. even if he had chosen ubuntu he would have found a way to blow his own dick off nobody takes him seriously though, right? its his schtick to implement technology in the most roundabout fashion possible and then watch proudly as it delivers 20% of what its really capable of.
|
# ¿ Nov 10, 2021 14:32 |
|
POSIX was such a lowest common denominator spec of what a UNIX implementation had to provide that IBM was able to build a POSIX compliant ( and UNIX branded maybe??? ) abi into MVS - os/390 - z/OS that should tell you all you need to know about the value of something being POSIX compliant.
|
# ¿ Nov 12, 2021 14:14 |
|
docs say its currently POSIX.2 compliant for shell, some utilities and kernel abi and i think thats just nasty
|
# ¿ Nov 12, 2021 15:53 |
|
BlankSystemDaemon posted:
i simply cannot innovate when you make me comply with these silly standards
|
# ¿ Nov 12, 2021 15:58 |
|
isn't coreOS like an appliance-ified linux with immutable-ish / , and config managed by crio when fed base64 encoded yaml somehow ? e: oop - ignition handles config, not crio - which is a podman/docker/container runtime fresh_cheese fucked around with this message at 22:44 on Feb 2, 2022 |
# ¿ Feb 2, 2022 22:40 |
|
bah. implement a key based record search with arbitrary block sizes in the disk hardware itself if youre really serious about that sort of thing.
|
# ¿ Mar 23, 2022 15:36 |
|
yummycheese posted:you jest but i imagine this type of discussion comes up frequently at the CEPH project. not a jest, s/360 did it in ‘64 with the Count Key Data disk io architecture ref: https://en.m.wikipedia.org/wiki/Count_key_data its literally “select all data records with key=<zipcode> “ implemented in the disk hardware/firmware its still there today in z/OS VSAM with any modern ECKD dasd. DB2 on Z is basically just a lock manager and sql parser on top of vsam data in eckd disk.
|
# ¿ Mar 23, 2022 19:14 |
|
the best laptop experience is to use it as a remote desktop viewer so you can close it and get more beer for a while as things happen elsewhere on teal computers mac/win/linux is kinda irrelevant imo
|
# ¿ Jul 28, 2023 22:39 |
|
mystes posted:g3 Mac im sorry i said a real computer
|
# ¿ Jul 28, 2023 22:42 |
|
oh. my bad. carry on then.
|
# ¿ Jul 28, 2023 22:49 |
|
would there be a productivity difference between teal vs tangerine or grape? i would think the red/strawberry would have to be fastest
|
# ¿ Jul 29, 2023 00:00 |
|
b0red posted:waypipe oh hey x forwarding sweet that was the thing i was waiting for before giving a poo poo about wayland depends on how it works tho. if the process keeps running remotely and you can get the ui window open again without losing state after sshing back in from a different client ip because youre in a different building then awesome! if the remote process dies when the ssh session drops then meh. vnc remote desktop is better for mobility.
|
# ¿ Jul 29, 2023 04:00 |
|
all the development ive ever done would run in literally any posix environment with probably an hour or twos effort. this should tell you all you need to know about both posix and my development skillset.
|
# ¿ Oct 3, 2023 13:07 |
|
oop - wait. theres a shitton of rexx in there and i dunno if orexx is fully posix portable. maybe not then…
|
# ¿ Oct 3, 2023 13:10 |
|
too bad python is utter poo poo
|
# ¿ Oct 3, 2023 17:21 |
|
load bearing whitespace is the stupidest poo poo ever
|
# ¿ Oct 3, 2023 17:35 |
|
god. i really am a relic of a bygone time. i stare at % all day on the terminal sessions i open first thing in the morning. i dont give a rats rear end about the differences between the different renderings of “-“ the only reason i picked up python was so i could write ansible modules to replace stuff i did with bash 20 years ago so the new kids could use it.
|
# ¿ Oct 17, 2023 12:42 |
|
hey - pretend im dumb: why do we need more than one “-“ ?
|
# ¿ Oct 17, 2023 16:29 |
|
yea see, i got bit by that “-“ is not “-“ poo poo last week and spent loving hours trying to figure out why my loving yaml was not parsing correctly. its because some “-“ opinion having motherfucker put a “-“ instead of a “-“ in the example documentation with the helpful ‘click here to select and copy this whole box to your paste buffer’ button and then, when i clicked it, and pasted it into a terminal, saved it, and tried to apply it DIDNT loving WORK keep your goddamned typography out of my loving examples user experience my rear end
|
# ¿ Oct 17, 2023 23:35 |
|
also gently caress whoever made smart “ a thing. do not loving change my “ to some other identical “ which also prevents example code from running. your user experience improvements are a piece of poo poo
|
# ¿ Oct 18, 2023 00:06 |
|
ASCII motherfucker! Do you speak it?!
|
# ¿ Oct 18, 2023 00:12 |
|
no, people who use em dashes in examples in digital form, and tools which allow em dashes when the rendered target is an example are the problem maybe if Courier only had the normal “ and - and everyone used it for examples thatd help?
|
# ¿ Oct 18, 2023 00:54 |
|
i prefer / be something of a fixed size that never grows after the initial os install is done. not even after adding ~200 packages after cloning the lun from the gold image to personalize the new system for its function. lvm / is gross and wrong because it opens a lot of exciting new failure modes that are really annoying to recover from. a standard / partition with lvm /var, /opt, /tmp, /home, AND /usr is the way. merged /usr into / is dumb.
|
# ¿ Oct 18, 2023 12:26 |
|
Athas posted:What is the value of never adding subdirectories to /? And why does the same not apply to /usr? add all the subdirs you want to / im not your mother? i would prefer they are their own fs mounts if theres gonna be more than a static config file in there. the point is to be able to boot the system into a state more capable than the initrd provides when lvm cant find all the physical volumes. if / is essentially static enough you can pretend its RO ish and recovery is a lot easier since you have more tools for diagnosis. /usr, /var, /opt, /tmp, and /home all grow as you add packages to a system. /bin, /lib, and /etc very much less so. honestly i really want pxe boot ro / with nfs everything else - olde diskless workstation style - just with san or iscsi instead of nfs.
|
# ¿ Oct 18, 2023 12:55 |
|
if lenny hadnt put all of systemd into /usr it wouldnt even be an issue. why TF did he do that anyway? just to be an rear end in a top hat and force the issue because of opinion stuff or was there actually a technical justification behind that decision?
|
# ¿ Oct 18, 2023 13:25 |
|
rescue media? you rely on booting rescue media? loooooool okay. i hope that continues to work for you.
|
# ¿ Oct 18, 2023 15:01 |
|
BattleMaster posted:
id be fine with that. all i really want is to be able to have a simple partition with a minimum system install on it, with an “everything else” lvm-like region that goes with it. everything you add post installation goes in the lvm managed area. im not emotionally attached to /usr , i just want to keep this distinction between the parts of the system that support the shell , devices, and file system management vs everything else. BattleMaster posted:
its probably obvious i dont really give a rats rear end about typography. im happy other people like typography because they make cool stuff, but cramming magic typography transformation logic into every possible rendering of text is a poo poo idea. magic quotes in a wysiwyg editor? awesome. magic quotes in a shell in xterm? noooooooooo bad plan.
|
# ¿ Oct 19, 2023 11:24 |
|
Antigravitas posted:It's more complicated, partly for hysterical reasons, partly because of Debian's commitment to making distribution upgrades happen via package manager. why the hell did everyone decide to merge / into /usr instead of the other way round? /usr was the hack, get rid of that, dont put everything in it?
|
# ¿ Oct 19, 2023 11:43 |
|
fuckit lets blow up the FHS the ideal file system structure: / is ephemeral and managed entirely by the kernel with some options to direct it to create specific symlinks to places /os/base is generally RO and consists of the executable code currently smeared across various bin and lib directories. /os/p1 , /os/p2…n are union mount overlays for /os/base : this is how patching is done /conf/base is RO config files as provided by things that come from /os/base /conf/local0 is a union mount over /conf/base for customization /conf/local1..n are additional union layers over that for additional customization /state is anything thats not code or config, also has union layers like conf to permit customization anything not part of the os goes into unique / subdirs by vendor/project/whatever you need it to be. the critical requirements are that they stick with the hard distinction between code/config/state and union based layers for patching. whether non os config goes into /conf or a different / subtree per subsystem is left as a point of contention for future lols poo poo. i reinvented containers again.
|
# ¿ Oct 28, 2023 15:29 |
|
didnt VMS kinda do that?
|
# ¿ Oct 29, 2023 20:27 |
|
z/OS has the VSAM access method which is also a key value store ref: https://www.redbooks.ibm.com/redbooks/pdfs/sg246105.pdf the real cool bit is VSAM rides on top of a disk architecture called ECKD ( extended count key data ) which is a disk hardware/firmware key value store, so its fast as all hell
|
# ¿ Oct 29, 2023 22:14 |
|
if you have only one backup you have 0 backups if all your backups are stored in the same place you have 0 backups if your backups are writable from the prod env you have 0 backups if you have not verified you can recover anything you might want from your backups, and do so regularly, you have 0 backups
|
# ¿ Nov 22, 2023 16:26 |
|
i been doing rsync but restic looks neat thx OP
|
# ¿ Nov 22, 2023 16:27 |
|
mystes posted:restic is cool and good but needs object store ah here we run up against my old enemy: my own cheap-ness fak. guess i oughtta look at some S3 costs anyway so i can me some get off prem backups
|
# ¿ Nov 22, 2023 16:39 |
|
mycophobia posted:actually it's one backup nah man the math works different for backups. theres like heisenberg uncertainty and quantum entanglement n poo poo.
|
# ¿ Nov 22, 2023 16:41 |
|
yea i had a whole presentation on the distinction between a copy and a backup. -> if ‘rm -rf /‘ or ‘lvremove’ or anything else within the system of concern can take out the backup, it aint a backup
|
# ¿ Nov 22, 2023 17:13 |
|
|
# ¿ May 13, 2024 00:30 |
|
mycophobia posted:i rsync to an external drive every once in a while when i remember and also feel like it. thats my backup did you verify the external drive is honest about its capacity and is all good blocks? maybe its lying that it stores anything past sector 1337.
|
# ¿ Nov 22, 2023 17:25 |