After decades in the trenches, this engineer is done with hype cycles
OPINION The real opponent of digital sovereignty is "enterprise IT" marketing, according to one Red Hat engineer who ranted entertainingly about the repeated waves of bullshit the industry hype cycle emits.
During a coffee break at this year's CentOS Connect conference, The Reg FOSS desk paused for a chat with a developer who was surprised but happy to find us there. We won't name them – we're sure that they'd prefer to keep their job rather than enjoy a moment of fame – but we much enjoyed their pithy summary of how IT has faced repeated waves of corporate bullshit for at least 15 years now, and how they keenly and enthusiastically anticipate a large-scale financial collapse bursting the AI bubble.
This vulture has been working in the tech field for some 38 years now, and the Linux developer we spoke with has been in the business nearly as long. We both agreed that the late 20th century – broadly, the period from the early 1990s onward for a decade or so – had mostly been one of fairly steady improvement. Then, they suggested, roughly following the 2008 credit crunch, we've had some 15 years of bullshit in tech.
They called out about half a dozen particular instances of what they considered to be bullshit technology. We were too busy laughing sympathetically to whip out a laptop to make notes, but as best as we can recall the sequence, they were:
Adding back some of the rather invective-laden commentary…
Containers: Sure, yes, they work, they are handy for testing. But they aren't a deployment method. You shouldn't need them. Anything that you can run in a container, you can just run on the bare metal, and if you're not competent enough to get – and keep – that working, then you probably aren't competent enough to deploy a container either.
Kubernetes: If you don't need containers, then you don't need another vastly more complicated tool to deploy those containers. The chances are, you are not a vast multinational that must be able to withstand ten million potential new customers visiting your site all at once. It won't happen, so you won't lose any of that imaginary business.
(This is sometimes known as the Use One Big Server approach, and in our humble opinion, it has great merit.)
The cloud: Nebulous by name and by nature. Who thought it was smart to take all your company's important data and hand it to some internet rando – probably the lowest bidder – trusting them to store the crown jewels, keep them safe, and never ever peek at them. If that sounds reasonable to you, maybe you should try selling homeopathy.
(This can be summarized as There Is No Cloud – There's Just Somebody Else's Computer, and was being spelled out in clear language in 2015.)
Anything "as a service" – it doesn't matter what: Infrastructure as a service – if you need servers, buy servers, or rent your own private servers. Nobody else will ever care as much about your servers as you will. Platform as a Service – now you don't even get servers, just OS instances. That's even worse. Software as a Service? Now you don't even know what the server is, or where it is, or what it's running; you don't get software, and you don't even know what data you have or how it's stored – you're paying for access to your own stuff.
(The problems with this entire concept arguably go back to Peter Deutsch's Fallacies of Network Computing some 30 years ago.)
The blockchain, and anything built on the blockchain: the world's slowest and most-distributed database. Cryptocurrencies? Hashcash on the blockchain. NFTs – URL shorteners on the blockchain, only they're longer rather than shorter. Worthless. Web3? Get ripped off, as a service.
Which brings us round to "generative AI" or, as we prefer to term them, large language models, powered by the transformer algorithm. If The Financial Times can explain how it works to a banker in a couple of thousand words and a few minutes, it can't be that complicated or hard to understand, and it isn't. It's predictive text turned up to 11. It can't even count. As Daniel Stenberg, author of curl, caustically observed:
(This vulture laid out some of his case against it when Gentoo and NetBSD banned LLM bot slop in 2024.)
The "i" in "LLM" stands for intelligence.
Honestly, we can't fault any of this reasoning. We've looked into the chronological sequence of the waves of marketing drivel, and it's not quite how we expected. Although the earliest mention of Salesforce.com we can find on The Reg is from 2002, when we called it "relatively new," it was founded in 1999. Perhaps the first mass SaaS offering to the general public was Google's Gmail in 2004.
Cloud computing in the sense of automatic creation and deployment of VMs arguably dates to Amazon taking Amazon Web Services live in 2002.
Whoever "Satoshi Nakamoto" is or was, their paper [PDF] introducing Bitcoin was published in 2008, although it didn't come to The Reg's attention until 2011.
2008 was also the year that the first version of LXC (you can still find version 0.1.0 on the downloads page) was released. Docker debuted in 2013 but your humble correspondent had predicted that Linux containers would be the Next Big Thing a couple of years earlier, back in 2011. We reckon we called it.
Kubernetes first appeared in 2014, although Google had been running "Borg" internally since around 2008. We still harbor a cynical suspicion that Mountain View threw it over the wall for no other reason than to distract the Penguinisti and keep them busy.
So in chronological order, those are:
1999: SaaS
2002: Cloud computing goes mainstream with AWS
2004: SaaS reaches the general public
2008: LXC frees application containers from FreeBSD's Jail and brings them to Linux… and Bitcoin delivers full industrial-scale mass-production of the ancient scam of the pyramid scheme
2014: Kubernetes is loosed upon an all-too-willing tech world
2022: ChatGPT opens to the public, or as The Reg calls it, another AI to fill the world with kinda-true stuff
All right, yes, rather more than 15 years. "A century of tech BS" seems a bit over the top when it's only 2026, but it certainly feels that long.
Obviously there are many more potential candidates, but we thought this was an excellent top six. Some of the other contenders are more niche, from the eternally awful Jira to the project managers' religion of Agile. Which of your bêtes noires did we miss? ®
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。