Dead ends of information technology | Hacker News Reader

Dead ends of information technology

github.com

46 pointspeterhunt70 comments

There's a lot to quibble with, here.

But I find this interesting/worthwhile overall because it demonstrates just how broad and deep the landscape of dead ends is.

Even Ansible is dead by now.

Extremely, absolutely, definitely no. It's alive and well, and still being used for new deployments.

There was a time when every startup used it. Now I don't know any startup that is. I suppose it's possible that the big infra managers use it with their teams. -- I wouldn't know, but the necessity for Ansible by the bulk of developers has evaporated due to containerization.

Ansible + Terraform is basically the modern infra stack if you’re not on k8s.

Ansible is a powertool for implementing idempotent configuration management. That pattern / paradigm is here to stay.

I'd never call anything imperative (as opposed to declarative) a power-tool for implementing idempotent anything.

Amen. Useful tool _in which one might_ implement idempotent patterns, yeah.

It pretty much is declarative enough to count, if you describe your operations in enough detail and rely on modules where you can. Not many people bother to go 100%, but you can easily get to 99%.

Mail.ru and Yandex used it, i've seen people deploy their Kotlin web apps

plenty of startups still using it

basically if you don’t want to go kubernetes, you’re gonna use ansible

I use it in my home setup to spin up some service that run in containers automatically. The flow is: change docker file, push, CI runs and validates, another CI runs that uses ansible to run those containers. I probably could connect directly to that docker machine but I don’t want to expose Docker API to the network so I just use ansible.

ansible is very much alive. kubespray deploys kubernetes clusters with ansible for example.

chef and puppet though? THOSE are dead…for new applications. many MANY big companies still use Puppet extensively for legacy config management.

Who (or what) is the contender to replace Ansible?

I think Nix is most poised.

If you don’t have a way to do immutable infrastructure, Ansible is still extremely alive.

Containerization in general?

Who configures and installs software on the container hosts? How do you deploy applications directly on bare metal or VMs? Those are still sometimes required or preferred for various reasons.

IMO, you can speak about the death of bare metal (which is slowly happening), but definitely not tools like Ansible. They've found their niche.

> If you don't need a network drive any more (since you use GSuite, Office 365 or Nextcloud), then it is likely that you don't need a VPN any more. Leaving a very skinny almost serverless LAN.

As content providers becomes more regionalized; I find VPNs becoming more (and not less) necessary.

A VPN also usually let's you bypass restriction on firewalls, especially with Wireguard protocol that is fairly new and it's over UDP.

Especially in shared Wi-Fi networks that means usually better performance, and less problems with rejected connections to whatever service, or certain type of traffic not permitted, DNS requests that are redirected, and so on.

I don't usually use VPN for privacy, since these days you have HTTPS that protects the communication, but they are super useful if you want to use a network that is not yours without issues.

Beside that these days we may not have a file server, but we have a ton more of devices that we want to access when we are not in our network, without relying on a cloud (and thus giving away our privacy): IP cameras are the perfect example, I don't want the video stream to pass trought the cloud, I can connect in VPN to my home network and access the video stream. Beside that there are a ton of "smart" devices inside your home these days.

> I don't usually use VPN for privacy, since these days you have HTTPS that protects the communication, but they are super useful if you want to use a network that is not yours without issues.

https doesn’t protect against URL snooping, so privacy is only as good as the request.

Not that VPN is a real solution, but I find the GP/source’s thought process on this as sloppy.

> Parsing text with regular expressions is like eating rubbish. In the 21 century, we send and receive data structures.

Of course you do, if you're starting from scratch. But since we tend to work on the Internet, which has a long history and myriad data formats and protocols, we often (rightly) use regex to validate and parse.

In the Unix philosophy, that means you have a series of tools that each deal with one of these protocols. If you do one job, do it well - and regex is a bad, if very quick, solution to structured data parsing problems.

> The bash is the default shell,

not on macos it is not.

> Emacs

I can assure you, the people who used emacs 20 years ago are still using emacs.

> Desktop office: MS-Office, Libreoffice. These programs were very important in the past. Today most people learned, that you can write the text directly into the mail body.

lolwut

> Nobody wants to call malloc() and free() any more.

um, embedded?

> Parsing text with regular expressions is like eating rubbish.

ok, I stopped here. This guy is a clown.

> I can assure you, the people who used emacs 20 years ago are still using emacs

And there's new users too! I for one am only about 20 years old :)

A lot of this is incredibly myopic. The author seems to think that because a tool isn't useful to him that it isn't used at all. Also, there's the repeated notion that some currently-popular new toy has "won" and forever displaced everything before it. Nothing really works that way.

Yep... "tech of the week has won", until next week, when "tech of that week has won", and most of the stuff declared 'dead' will still be working and in use pretty much everywhere.

C dead? XML dead? GUIs dead? Regex dead? .... yeah... no.

Cobol, so dead. Perl, dead. Remember when PHP was dead? And there were two things on the front page today about PHP ;). No death, only zombies.

ha! I'm outraged that RSS doesn't make a mention on this page... it seems to be dead every week or two here :-)

Sarcasm much huh?

I reached this post from an rss reader

I think that's exactly his point.

yeah it was.

Especially given how contentious some of his points are.. RSS really does belong on that page...

Surprisingly microkernels seem to be coming back. Google has Fuschia, Facebook has Reality OS.

It a lot about the control. Big companies have noticed that they can’t get everything they want into upstream.

Mobile is the future - Mobile code, the interface it runs on will be local, not served via the strings and tin cans that is http(s). Code written by someone else, in whatever language, using whatever methodology, no matter how unsafe, needs to be run in a secure manner, given only the capabilities required to do the job. The OS needs to default to NO access, and stop malicious code, and confused deputies, from breaking things.

If your OS can't run a chunk of code that uses GUI gadgets directly, you don't get to play. The desktop is back, but not really the way you think. The metaphor of directly manipulating UI bits with code is what we get to keep. Direct access to the file system is right out.

Linux won't get to play, because it can't run mobile code without giving away the store. Neither can Windows, MacOS, iOS, Android, etc.

This is why we have an interest in Microkernels.

Hindsight is 20/20.

Nuanced toolbox for the real world > absolutist myopia.

Whenever someone claims years of experience, they've given up on mastery.

Native produces better, faster apps than introducing abstraction layers. Electron and Atom are resource hogs that force you into wasting resources. Appcelerator, Phone Gap, Gtk+, and Qt are for when you don't have the time or money to make something quality.

Saying NFS isn't used anymore is ridiculous. Modern NFS 4.2 with pNFS features is badass.

> Electron and Atom are resource hogs that force you into wasting resources

You can repeat it as much as you want but poke any modern laptop, how many native apps and how many electron apps are there? Spotify, VSCode, bunch of note taking stuff, trello for desktop? And those that are not electron-based, they are simply running in browser. In other words - who cares what you think about performance?

> Saying NFS isn't used anymore is ridiculous. Modern NFS 4.2 with pNFS features is badass.

Again, nobody cares that you think it is badass. My work is still on NFS 3-something because NFS 4 introduced performance issues and restrictions that make it impossible to use. And you know the way to upgrade? Yeah, get rid of NFS and develop a cluster of independent agents that talk over grpc.

> Again, nobody cares that you think it is badass. My work is still on NFS 3-something because NFS 4 introduced performance issues and restrictions that make it impossible to use. And you know the way to upgrade? Yeah, get rid of NFS and develop a cluster of independent agents that talk over grpc.

You don't know what you're talking about. Grpc isn't a distributed, fault-tolerant filesystem, buddy. Maybe you should be barking up the Lustre or GPFS trees if you actually did HPCC in the real world as I have. You may sit down. Mic drop.

Yeah, and as the user of an older, "underpowered" device, I'm excluded from all these "modern" apps because they either freeze up as a desktop app or freeze up as a JS app with no fallback, and I'm unable to participate in many of the groups which use them for communicating.

But, of course, who cares about me, I should just buy a new computer, right?

> But, of course, who cares about me, I should just buy a new computer, right?

Unfortunately it is not only you. _A lot_ of my employer's customers complain about frontend performance and the fact that the UI is generally made with assumption of fairly large display (almost everyone in the company has got 27"+ monitor whilst our customers are often on 2-3 years old business-versions of Lenovo or something similar with 14 inch of screen and 8gb of RAM top).

And yes, that's right, who cares. People like you are minority (people on phones are majority), people in enterprise settings don't have a voice and choice, and economy drives everyone towards using web stacks.

Also I guess my expression wasn't exactly clear. What I'm trying to convey is that technical superiority doesn't matter and users don't care about technical superiority. They care that music plays right away when they tap "play". And it is way easier to deliver that experience using web stack than any kind of native stuff.

> 2-3 years old business-versions of Lenovo or something similar with 14 inch of screen and 8gb of RAM top).

I have a 4 yo Lenovo 14" T480 with 64 GiB of Samsung RAM, a magnesium top, 4 TiB SSD, quad core 8650i with discrete graphics, internal WiFi 6E, WWAN, and real 10 hour battery life with hot-swappable 72 Wh secondary batteries and an external battery charger. Newer models like the T490 and T14 G1/2 are much less usable, serviceable, and upgradeable. I have a whole stockpile of repair parts so I'm keeping this for years and shunning developer waste.

>People like you are minority

Disregard for minorities is generally frowned upon by human societies, isn't it?

>And it is way easier to deliver that experience using web stack than any kind of native stuff.

I think it's totally doable to build something performant using a web-based "stack" if the stack itself doesn't weigh more than five elephants.

> Disregard for minorities is generally frowned upon by human societies, isn't it?

Not really, it is relatively recent development in a limited set of cultures/subcultures.

And even in those cultures, it is not just any minority but a marginalised minority.

> I think it's totally doable to build something performant

Again, no one cares, they are performant enough to deliver experiences they want to deliver and there’s no incentive to do anything “better”. Better in quotes because “better” depends on priorities.

Just to clarify, I’m not saying it is good that we’re in that situation. I’m saying that reality is what it is and that I don’t see any stimuli for it to change in near future. Even death of Moore’s law didn’t help

Are native GUIs really a dead end?

Not on Apple platforms. The article links a "gtk and qt trend at Stackoverflow". Adding swiftui and uikit to that trend: http://sotagtrends.com/?tags=gtk+qt+swiftui+uikit&relative=t...

Unfortunately, it seems that way. If it isn't dead then it's on the endangered species list. Electron / web-as-native has clearly won and it seems like no one is even attempting to improve the state of native GUI development as a result.

> no one is even attempting to improve the state of native GUI development as a result

SwiftUI, React Native, Flutter disagree with you

I am not so sure about React Native. Like Electron, it is framework for people with web developer skill, missing a lot what real native frameworks can do.

> it seems like no one is even attempting to improve the state of native GUI development as a result.

Who do you include as “no one”? Apple for example is going full steam ahead with its native GUI frameworks.

Aye, but in the broader landscape, they're a bit player. The average person's exposure to Apple GUIs will be on their phones.

Not on mobile...

How many of the most-popular non-game mobile apps are just shells around a mostly web interface. I can think of quite a few.

Choose 2: Born in 1976, 43 years old (at time of writing), document was updated 2021-Oct-29

It was written in 2019. He even mentions that at a few points.

> Today in 2019...

> Now, in the year 2019...

> Up to now (2019)...

Correction: it's been continually updated since 2019. I'm not sure how to judge that one.

But given the uselessness of his git commit messages, I'm not inclined to judge leniently.

If you click on blame, it shows that he wrote '43 years old' 3 years ago and hasn't edited that line since.

https://github.com/guettli/deadends-of-it/blame/master/READM...

> today NFS (Network File System) in a PC LAN does not make much sense any more.

> Today people either use SMB to access files on a network share, or they use a web-based file service

What? Properly configured NFS is blazing fast. And why would I use a cloud if I have some files locally?

> The winner takes it all: Linux

FreeBSD is still used by one of the largest CDNs, Netflix.

> If I would be one of them, then I would try to find a new job with a better prospects for the future.

Yeah, automate that with a web-based frontend.

I don't agree with almost anything this guy says. You can't derive general laws from just one person's experience.

> There are still a lot of NFS based solution for server-to-server communication, but I would use a Ceph-based solution, if I could start from scratch.

srsly? as if the two have anything in common and the same complexity?

They solve all the same operational problems — how do I access these files on multiple machines but NFS is strictly worse in pretty much every way to GlusterFS and Ceph.

NFSs HA story is almost nonexistent and comes with list of buts that’s longer than the positives. As a technology it’s dead for anything other than the tiniest of internal IT use cases.

Exactly. NFS takes a couple of lines in /etc/exports. Ceph is large, scary, and multi-node.

> It is very uncommon to automate MS-Word or Excel via COM these days. I am happy.

Yeah, but I have a few bazillion lines of .bas that I'm not afraid to dip into on occasion for purely internal reasons.

Microkernels are necessary if you ever hope to secure a machine exposed to the internet. Linux can't be made secure.

However, with the knowledge that DRAM isn't secure thanks to rowhammer, etc... nothing will ever be secure again, unless it uses Static Ram, which is too damned expensive.

Computers are just going to get less and less usable as everyone forces the user interface into smartphones.

>https exists. Why a second security layer? Of course there are a lot of use-cases where you need more than one layer. But for most cases https is enough.

Https isn't everywhere you think it is. (e.g. Linux and mobile):

https://transparencyreport.google.com/https/overview?hl=en

HTTPS also locks out access in a whole slew of scenarios.

Native Android apps perform much better than web apps.

Isn't this just saying yesterday's fads have been replaced by today's fads and you should use today's fads instead.

Kinda scared, I didn't know a lot of those old things, since I'm kinda young and I wasn't into computers.

Don’t worry, 10-20 years from now, you’ll know all of the things folks will be proclaiming as dead, and younger folks will tell you they never heard of most of them :-)

I disagree with some of the things on that list but generally I like the list.

Relying on google trends is the dumbest things I can think of

Almost everything declines over time on google trends

PostgreSQL is hotter than it's ever been, and is "declining"

Absolutely this. The past is not a guide to the future, at least in this sense. A lack of buzz does not mean something is dying; more often, it means that it is mature and well-understood, with much of the necessary supporting information already out there and readily discoverable.

It’s useful to summarise the leading solutions at a moment in time, but nothing ‘has won’ - it is only ‘winning for now’. That is the insight/hindsight that is missing from this borderline-obnoxious opinion piece.

Reply on news.ycombinator.com