Should we throw current software in the trash and start fresh?

Compilers ignore comments.

That guy constantly echoes my own opinions on programming.

We're dealing with unimaginable amounts of terrible design and implementation here. There's no other industry where utter fuckups are so commonplace and tolerated. I think it does all need to be thrown in the trash and the fundamentals (including hardware) replaced by something very pared-down, distilled, straightforward, and opinionated. But of course it'll never happen.

Who

casey muratori

Gnome should start fresh.

Attached: e86.jpg (677x1400, 199K)

icewm is better than xfwm

I had this feel even during my CS degree
it all seemed like a house of cards, poorly designed, irregular sized cards

it would be nice to have a fresh lean secure network stack and os without all the obsolete compatibility crud

email needs a redesign from scratch as well

EXWM is better than either of those

One of the issues is that we are running our OSes on many different platforms.

If we go back to the Atari/Amiga days, an OS was designed for a very specific hardware subset and allowed much cleaner, targeted code.

youtube.com/watch?v=kZRE7HIO3vk
good video, sums up my feel exactly

I think we're going to see some efforts at getting back to basics with open source hardware, security is more important now than ever and software is becoming annoying to use

what are some more tech youtube channels like this?

How do we fix it?

LOC is not an argument.

Not that guy (first time posting in thread). There are very few, but this guy is one of the best imo.

FYI there are a few historical reasons gnu utilities all have a bunch of features and a lot of extra code:

1) to avoid infringing on proprietary unix copyrights
2) to make them more appealing to people on older proprietary unix systems that probably had a lot less features
3) because anyone who wanted could contribute, they would accept patches from anyone who wanted to write one

unfortunately feature creep is a natural part of any popular system, no idea how to fix this without just rewriting stuff every so often. also i do like plan9 but it isn't really practical for most users, it's a research OS

now compare performance between the two

The Plan 9 implementation is probably faster than the GNU one. The one real slowdown I can see in Plan 9's implementation is the small buffer size. but increasing that to 65535 or something is absolutely trivial.

Left isn't even POSIX compliant.

No, it's probably not. Read the code, the GNU one makes a ton of optimizations, such as finding the optimal blocksize.

Swap kde with xfce

Why should it be?

this

>Why should my OS support existing software?
This is how you end up dead like Plan 9

Why should it?
Why not just make something better?

The point of Plan 9 was to be a research operating system. You never saw Research UNIX from Bell Labs take off commercially, did you? No. You saw Version 7 UNIX and SysV UNIX take off.

Plan 9 is a beautiful operating system all things considered, exactly because it did away with a lot of POSIX cruft.

The realization of how complex that operation is. Is disgusting me.

Both are great but Xfce is far light and usable.

Hardware will eventually get a point where it can't slim down anymore or get any faster, but the bloat that the modern software is built upon will only get larger and larger. If we don't adopt a solution like this, programs will slow to a crawl and end up running slower than any other point in history.

the display should be 640x480, 16 bit color. unsigned ascii. image comments in source code

Multiple people have posted their observations about less loc translating into better projects, regardless of ANYTHING else. Language, editor, etc. Linecount was one of the most important apsects on its own. Fewer directly means fewer mistakes and fewer instructions, at multiple levels.

At the end of the day, despite it seeming like the case, "programming a machine" is not exactly what we're doing. We're writing text and compiling it to bytes. We read and write in a language made for humans, to control the machine. And what matters then is how people interact with text, not how machines work, or anything like that.

>Or will we just end up with hacked together code all over again?
Yup. Human nature. Everything in society is hacked together shit. We're not smart enough to create anything cohesively in the long run.

Because people don't fucking WANT better you mung. Stop trying to ignore the last half-a-fucking-century of computer science.
We had a dozen options that weren't based on the 8080. Yet here we fucking are. Why?
Because when you tell a banker their old-enough-to-vote line printers aren't fucking supported anymore and the COBOL they use as no living developers they LOSE THEIR FUCKING MINDS.
There are places of learning near me that I know for a fact are still running Novell.
Companies don't want _good_
They want compatible and cheap.
And companies drive development.

I don't know who you're trying to convince. In any case, you can't stop progress.

yes and yes, desu

Companies may want cheap but they don't get it. The amount of resource wastage that is a consequence of software fucked-upness is incredible.

>such as finding the optimal blocksize
Who cares? It's cat. Also the optimal blocksize for dealing with modern-day nonvolatile storage is 4K, end of discussion.

You entirely miss the point of developing research operating systems. Plan 9 does actually have a compatibility layer for crusty POSIX software anyway, called APE. It's all quarantined there.

It's practical for clear-sky projects.

hacked up code will only get worse because its a function of the dynamism and speed of the economy vs the speed of programming.

if there is a total collapse of the programming sector then code will be engineered instead of hacked together again, until it heats up again.
but right now ideas, investor capital, and being first to market is most important compared to anything else

the right side is better though
reddit.com/r/unix/comments/6gxduc/how_is_gnu_yes_so_fast/
there's a reason GNU 'yes' can do 10 gigs a second, vs net bsd 100MiB

>There's no other industry where utter fuckups are so commonplace and tolerated
you sound like a gaymer complaining about video game journalism
you're naive, every industry in the world is full of fuckups, this is just the only one you know

What's the aerospace equivalent of an Electron app that can't keep up with human typing? Or the structural engineering one? Or any other fucking field.

bad example
any field where if you fuck up, people die, there won't be many fuck ups
any field where if you fuck up, people are slightly inconvenienced, there will be as many fuck ups as you can imagine

if your social app with network effects gets to market 6 months before your competitors, but your one uses lots of ram and is slow then you still win and get a billion dollars

Compiler ignores everything that is not code so its irrelevant.

But, there's no reason to have 10 gbps of yes

what do you want, less LOC, or faster software?
protip: they're not synonymous

Not that user but simplicity should win out in the end. Simple programs are more portable. The only place where this becomes a problem is math libraries and other high performance software.

Attached: 1536279994234.jpg (1117x1600, 1.04M)

"durr I dont know what options are"

just use openbsd newfag

xfce is fucking garbage, even worse than gnome. they belong side by side

all we need is just one strong ruler.

Fixed.

Attached: 1536711931216.jpg (677x1400, 219K)

HARMFUL BLOAT

>but when the world needed him most, he vanished

Attached: 1475071300542.jpg (1148x917, 155K)

pretty fucking useless options don't you think? and why are they in cat?