Linux: vim isn't moe

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > Linux: vim isn't moe

«‹›4 »

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

mods namechange me to maximum Linux tia

# ¿ Mar 28, 2014 21:10

Adbot: ADBOT LOVES YOU

# ¿ Apr 26, 2024 00:28

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Mr Dog posted:

Kay Sievers is soooooo butthurt about being called out by Linus lol

it's like dude you hosed up quit being a child about it and move on with ur life

Nothing I've read about Kay makes him look good at all.

Lennart in comparison seems much better.

# ¿ Apr 10, 2014 17:23

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Mr Dog posted:

why do you need filesystem namespaces to host multiple services?
why do you need PID namespaces for gently caress's sake??

if a physical or logical co-hosting of multiple independent OS instances seems like a solution to your problem then it means that your operating system is a piece of poo poo to begin with, so loving fix it instead of stuffing it under the floorboards (i.e. adding another layer of abstraction like a lazy piece of poo poo instead of fixing the existing broken abstraction)

i think the issue is that some people use suse, some use rhel/centos, some are complete dumbshits and use ubuntu. some of these distros use rpm, others use deb, systemd unified a lot of the pointless bullshit differences between them and they're fairly similar now except for what package manager they use and what the details of their release policy and acceptable licenses in their repositories look like, but um, well, each of those three kinda have a company that wants to continue to exist and make money behind them, so there's still no unified linux platform out there to standardise on.

so we just pick one of suse/rhel/ubuntu, throw its entire humongous half a gig bag of bloat into a container image, and then put our 200kb of application code on top of that, forward ports into this thing and then call it a day

i.e. the problem is a political one, not a technical one, otherwise we'd all have a de-facto standard around idk debian or something (you yourself seemed to imply that yum and rpm are terrible poo poo being kept from being fixed by means of political backstabbing)

docker is literally just a frontend for kernel code virtualizing the pid namespace and the fs namespace

virtual cpus = threads, virtual memory = processes

imo cgroups should have been there from the start every process should run only in its little sandbox unless authorized to communicate with something else

# ¿ May 2, 2014 16:42

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

also vm hosts having page deduplication ends up working kind of ok

mirage OS is even more awesome, just compile a custom kernel with ur app!!

# ¿ May 2, 2014 16:48

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

ii oh el posted:

tex is just a festering boil that has been allowed to ferment for forty years and decided to evacuate itself all over your package manager

# ¿ May 7, 2014 18:07

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

osx is the worst of all worlds

it's a proprietary clusterfuck like windows, but it's a really lovely and terrible unix from the 1980s. i guess if all you want to do is check your stebemail and masturbate to daring fireball blogposts it is probably ok

if you wanted a unix desktop, to write software you will later deploy to real non-joke unix, it is a bad choice

lmao

bsd you're out of your league here

# ¿ May 26, 2014 21:55

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

osx is a really bad unix

you might like it for other things but holy poo poo is it a terrible unix

its a certified unix idk what else you want

# ¿ May 26, 2014 22:15

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

linux. shitloads of linux. linux at home, linux at work, linux in prod

i guess i have some osx taint on my phone and my htpc but i am not really concerned with how good the unix is on my phone

this explains a lot

# ¿ May 26, 2014 22:16

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

true mach is weird but in my perfect world we'd be on L4 style microkernels anyway

rip Jochen Liedtke

# ¿ May 26, 2014 22:19

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

the fact that osx was able to pass SUSv3 is the final word on how loving bad and useless SUS is

i'm not really seeing what's so bad about osx as a unix

works4me and other non sperglords

# ¿ May 26, 2014 22:43

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

ancient package managers >>>> no package manager

also modern package managers look pretty much like SysV packaging from the early 90s. as it turns out there just aren't that many ways to skin a cat.

re: X11 -- all my software is X11, and I want my software to work. I give less than zero fucks about how you feel about X11 technically. it could be the ugliest god drat thing ever written and it wouldn't matter a bit. not having good X11 support makes your unix a lovely unix

brew is a better package manager than most linux ones since it has a policy of not making GBS threads up the system + it delegates to language managers rather than halfassedly attempting to replicate them

mac app store for all of your consumer needs

X11.app is there for any idiotic legacy software you wish to poo poo up osx with.

# ¿ May 27, 2014 19:16

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

no.

you're a smart guy but you're hardly the first smart guy to conclude X should be supplanted by a compatible window server with a better API

known as quartz, with a cool api known as Cocoa!!

# ¿ May 27, 2014 19:17

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

x11 is pisstrash for idiots

icccm is the worst.

# ¿ May 27, 2014 19:18

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

brew is laughable horseshit. compiling on the target system is the wrong way to do pretty much everything. gems, pypi, cpan suck dead weasels through a hose.

gently caress that 80s bullshit, i might as well be trapped back in "./configure; make; make install"

p.s. "not making GBS threads up the system" is the point of a package system.

cool thats why brew now has binary packages that download /but also lets you build custom packages when u need to while maintaining hygiene/

# ¿ May 27, 2014 20:46

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

brew is laughable horseshit. compiling on the target system is the wrong way to do pretty much everything. gems, pypi, cpan suck dead weasels through a hose.

gently caress that 80s bullshit, i might as well be trapped back in "./configure; make; make install"

p.s. "not making GBS threads up the system" is the point of a package system.

welp the only package manager that i know that has transactional builds and package installation is nix which no one uses

everything else is various shades of making GBS threads up the system. brew is kind enough to let you blow away everything in its prefix and be ok

its not the best but its better than most distros who try to package everything under the sun and fail horribly

# ¿ May 27, 2014 20:49

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

It's almost like you guys have discovered that Notorious b.s.d. doesn't have any clue what he's talking about!

bsd is slightly more interesting to argue against than watching my builds percolate through the CI pipeline

# ¿ May 27, 2014 20:49

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Notorious b.s.d. posted:

so you consider it a virtue that brew can't manage system packages or patchsets? lol

yeah b/c apple does a better job than u or i can + mac app store infrastructure handles system updates

i would rather carve out my little dev env and then use vagrant/chef/puppet/ps DSC to set up a server with what i need tyvm

# ¿ May 27, 2014 21:12

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Mr Dog posted:

bsd are you in the most literal possible sense a greybeard

greybeard implies some sort of competence cloaked in knowing condescension

bsd displays neither

# ¿ May 27, 2014 21:13

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

That's not true. There's certainly a lot of condescension from Mr. b.s.d.

it was more the "knowing" part

# ¿ May 27, 2014 21:21

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Bloody posted:

this is incredibly readable, dunno why youd poo poo on it

the ink:data ratio is too high. bunch of extraneous ornamentation on the window borders

pointless "etching" effect

non anti-aliased text, icons

etc. etc.

but if you fix that it's pretty good, and it's pretty good for its time i guess

# ¿ May 28, 2014 21:12

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

No, I've certainly heard the term "network transparency" before, and sometimes by coworkers. It still doesn't describe any set of achievable goals, nor why traditional remote desktop solutions like RDP and VNC don't count, but extremely similar protocols like NX do.

theres a video by a the dudes behind wayland where he basically explains that x11 is awful at being network transparent

also lol@people complaining about plan9 being naive about network transparency and then clinging to x11 lovely network transparency

# ¿ Jul 16, 2014 23:52

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

No, I've certainly heard the term "network transparency" before, and sometimes by coworkers. It still doesn't describe any set of achievable goals, nor why traditional remote desktop solutions like RDP and VNC don't count, but extremely similar protocols like NX do.

b/c vnc is awful, rdp is a lot better but still sucks, and sperglords still think that X11 is fast because they used it over lan links and assume its fast over wan links

there isn't a good solution out there.

# ¿ Jul 16, 2014 23:58

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

ShadowHawk posted:

It doesn't break on Ubuntu out of the box since we install the Recommends by default. "Recommends" is a pretty strong Recommends (eg if a Recommends is uninstallable at package install time apt will throw an error rather than just charge ahead -- it's more like "depends that are able to be manually removed if you insist".)

If you're asking "why not make all defaults mandatory for everyone" then, well, there are different use cases. For Wine as well as any other package.

yeah then make them reqs and if idiots want to remove them they can use --force on dpkg or something

stop optimizing for edge cases

# ¿ Jul 17, 2014 09:32

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Sniep posted:

lol good luck with centralized logging when you have an actual large network

how bigs ur network

# ¿ Jul 23, 2014 16:17

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Sniep posted:

that's not our business. we're not going to invest the kind of money to make this a reality when things are just fine how they are. there is centralized logging from one angle but not full system logs because why? you can just get them from the system in questino.

lmao

# ¿ Jul 23, 2014 16:17

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

Thanks! I'm really excited here and I can't wait to show you guys what I'm going to be working on. It's a stealth mode startup but it's well funded and had a solid business strategy. They're already making money, in fact. The product is even on store shelves right now.

I'm going to be maintaining mutter because the company has a fork of it for their own OS. Whether I'm going to be doing upstream work first, I'm not sure yet.

GUADEC videos might be going up at some point, but the audio is probably bad. The guys that did the videos last year couldn't get a visa so we just had camcorders and the on cam mic.

lwn.net covered my talk though, so you can see coverage of it there.

yo u gonna keep doing xplain cause that poo poo is gold

also get monty to do more videos

# ¿ Aug 1, 2014 16:56

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

A file is a really cool data structure: a stream of continuous bytes, of variable size. Files are stored in filesystems, which have a hierarchical directory structure. It beats the crap out of the days when you had to statically allocate records of data and keep track of that, because there was no "filesystem" layer providing some virtual mapping between your storage medium and the locations of files.

This was why "Everything Is Just A File" became a meme and a big selling point: it took the programmers at the time a long time to adjust to the reality that they didn't need to be concerned with the storage medium, the layout of it, "records" or anything like that.

The unfortunate part is that a "a stream of continuous bytes, of variable size" is not a data structure that is suitable for everything. A quite large amount of things. But it shouldn't be discounted forever.

well, they are streams of bytes that appear to be continuous, i mean sparse files are pretty great for certain things

# ¿ Aug 12, 2014 18:35

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

it turns out bytestreams are pretty universal

# ¿ Aug 12, 2014 18:36

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

the filesystem metaphor, not really

# ¿ Aug 12, 2014 18:36

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

but it fits in many places and is not a bad thing

/proc is ok honestly, the alternative is a bunch of kernel syscalls and that's just awful given that C has no namespaces

so using the filesystem as a namespacing thing is probably all its good for

# ¿ Aug 12, 2014 18:39

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Mr Dog posted:

memfd syscalls getting merged, kernel secure rng syscalls also getting merged

(a memfd is a chunk of shared memory with an fd attached to it, as the name implies. the fd lets you pass the shm around more neatly using IPC. it can also be "sealed" to irreversibly make it read-only, which is useful when your're using this mechanism for IPC with an untrusted process)

so much for ~*everything is a file*~ (except when it is nothing like a file and just gets awkwardly crowbarred into the metaphor with implied semantics everywhere). perhaps one day we can get rid of /proc and /sys too and kill this garbage for good (/dev is never ever going away)

perhaps u would like WinNT,mr dog, it decided to go the way of constructing APIs for everything

# ¿ Aug 12, 2014 18:40

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

So universal that we bypass them entirely for device nodes and use a side-channel called "ioctl"s instead!

yeah and it sucks. ioctls are awful

sure i want a loosely typed api that i can easily gently caress up over writing bits to various pseudo-files, sure thats a great idea.

it;s a dumping ground for various crap

# ¿ Aug 12, 2014 21:02

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

Yes, and good APIs make great systems. I would much prefer a struct and syscall API over a kludge code that parses /proc/meminfo. I would love it if there was a way to programmatically add a user to a system besides forking out to useradd (adduser on Debian, because Debian policy dictates you need to use this Debian script that doesn't exist on any other system).

/proc and /proc/sys are a mix of system information designed for admins, APIs for developers, and configurable settings, with no thought put towards organizing or separating it. If you aren't careful, you can easily end up in the bad part of town and mess with something you shouldn't have.

this is orthogonal to pseudo file systems being the interface, honestly: it's very easy to write an awful ioctl api

like the whole poo poo with "rest-y" api's: it's bad but the alternative is worse

# ¿ Aug 12, 2014 21:04

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

Yes, as I said in the last post, I hate streams of bytes APIs and /proc too.

we are at an impasse

personally open/write/close is nicer than open/ioctl(fd,MAGIC_NUMBER_LOL,data)/close and friends

but i guess if u rly hate urself.

# ¿ Aug 12, 2014 23:28

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

pram posted:

saying you can deploy anything more complex than a database and webserver with no maintenance is not meaningful. you dont know poo poo ur loving dumb as hell stop posting

ya lol even on paas u need people doing maintenance

# ¿ Aug 16, 2014 23:06

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Suspicious Dish posted:

Nope, and it will likely never be. Most of the main developers are trying to convince corporate to just use Wayland instead, to no success.

lmao

So why exactly is corporate pushing Mir again?

# ¿ Aug 18, 2014 11:24

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

pseudorandom name posted:

serious post: running a VT100 emulator in kernel mode makes about as much sense as rendering TrueType fonts in kernel mode

significant portions of win32 gui were brought into kernel mode for performance reasons

# ¿ Aug 30, 2014 11:50

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Cocoa Crispies posted:

the brazilians i know with apple products seemed pretty cool, but they could afford to make trips to the us to buy 'em lol

if you're ever on a flight from MIA to GRU or GIG everything about peoples' checked luggage is hilarious: the worst kind of baggage completely overstuffed with stuff that's cheap here

blame the import tariffs

Rich Brazilians will go to London (lol) because apple stuff is cheaper in the UK

# ¿ Sep 1, 2014 12:17

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

Mr Dog posted:

anyway kdbus and gnome sandboxes will solve this for the narrow case of desktop applications (which tbh nobody really cares about for linux anyway). so you'll have a sandboxed GNOME Weather applet and GNOME Music application that can be released to users directly via self-contained release ZIPs from upstream downloaded via the GNOME Software application. Great.

app stores don't really work for server applications though, which is why I'm not so keen on this one-size-fits-all solution. There the problem is slightly different: all of the languages used to develop server apps have their own siloed package managers which don't interact with rpm or deb at all (node.js npm, Java maven, Perl CPAN, whatever thing Python has). But arguably that isn't important as long as you wall off some chunk of the FHS for them to play in, then set up ansible/puppet/whatever to use the system-level package manager to install your db servers and managed code VMs, then in parallel command the managed code VM package managers to install your application packages. I dunno. It doesn't seem like a problem that really needs fixing.

The only real area where this becomes a problem is when you have rpm/deb-packaged desktop apps written in these managed langs. Then you end up making rpm/deb packages for CPAN modules or whatever and it all goes straight to hell.

Either way these are two very different problems with very different desired final solutions. Poettering's Hall of Mirrors doesn't really seem to solve either of them particularly well.

NixOS solved this problem

hermetic builds are gr8

# ¿ Sep 1, 2014 18:30

Adbot: ADBOT LOVES YOU

# ¿ Apr 26, 2024 00:28

Malcolm XML: Aug 8, 2009; I always knew it would end like ｔｈｉｓ．

BobHoward posted:

no. they have an aes-xts fde layer in their os which works on any block device ever made (not just ssds), doesn't depend on firmware written by the storage industry to not have security flaws, and has a number of other advantages such as being architected to allow your regular user password to unlock the drive, allow multiple user accounts' passwords to unlock the drive, allow them to offer an optional service where an extra unlock key is generated and stored with apple for disaster recovery if you forget your password, and more. (the disaster recovery feature is for users who are interested in encrypting to protect their laptop from petty theft rather than obvs)

their fde uses the aes-ni acceleration block integrated into all modern intel cpus. it can keep up with the fastest ssds apple is shipping without chewing much cpu time

literally the only downside is that it doesn't push the encryption out to the disk to satisfy spergs

actually the real only downside is that sandforce controllers want to be able to compress data to hit their max performance and lol if u think the drive can compress aes encrypted data. but also lol if u think apple sources very many ssds with a sandforce controller

FDE is best handled by the drive so it can do it directly in hardware for power savings

ms has bitlocker

# ¿ Sep 6, 2014 20:24

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > Linux: vim isn't moe

«‹›4 »