I am genuinely interested in hearing why we collectively ditched XMPP. I would love to hear someone who has been in the weeds on the development or even just following closely.
Edit: Seems someone beat me to it with a good reply.
Depends if you mean just the technology or using it in the small federated spirit. Google Talk and Facebook Messenger were XMPP all the way through and worked with vanilla XMPP clients. Slack wasn't XMPP but supported it via a gateway until it was dropped.
Not sure how popular the small federation was back then, but I know Mac OS X Server touted an XMPP server and that was a first-class feature of iChat.
> Google Talk and Facebook Messenger were XMPP all the way through and worked with vanilla XMPP clients
I remember this, it was great to connect to absolutely every chat platform with bitlbee and pretend that all my chats were just DMs on some irc server somewhere
Likely you do or have without knowing it. The protocol is used in telecom quite a bit for all sorts of things. Jitsi is built on XMPP. Lots of games use it for chat - league of legends and unreal engine I believe. Xmpp shows up in all sorts of places if you look.
XMPP had rather bad name. Well-known design issues causing message losses, fractioned ecosystem due to varying implementation of extensions, unsuitability for mobile clients, absence of synchronization between clients, absence of end-to-end encryption. Most of these issues were (much) later fixed by extensions, but Matrix (or Signal for those who do not require federated one) was already there, offering E2EE by default.
Even today, E2EE in XMPP is rather inconvenient compared to Matrix due to absence of chain-of-trust in key management.
Sometimes I wonder if the endgame is each person having their own XMPP server for their set of devices. S2S is your E2EE then. Your chain of trust is your existing CA, unlike Matrix which starts from scratch. Cause XMPP wasn't designed from the start for clients not to trust servers, plus the fragmentation of C2S extensions was always a pain.
It's not a bad solution if someone can make it easy, even if it's a managed service that just lets tech-savvy users export it to self-hosting if they want.
First-class integration with two of the world's largest social networks put XMPP in practically everyone's hands for a time, but when all the major hosts left, network discoverability and typical account longevity dropped drastically. The landscape is bleak today.
And since then, our collective needs and expectations of a chat platform have expanded. XEPs have been developed to bolt much of that functionality onto the base protocol, but that has led to a fragmentation problem on top of the bleak server landscape.
This unfortunate situation might be navigable by a typical HN user, and perhaps we could guide a few friends and family members and promise to keep a server running for them, but I think the chances of most people succeeding with it are pretty slim today.
Facebook never had "first-class integration". It was just a client bridge - you could login into Facebook Chat using your XMPP client, but it was a completely separate network, unlike Google Talk which was an actual federating XMPP server.
(And my point regarding support on Google/Facebook was that their users could chat with me over XMPP without having to leave their familiar sites, sign up for anything new, or do anything else special. That put it in easy reach of the masses.)
The same could be said about various XMPP transports that I've used back in the day with Google Talk to access all sorts of IM networks. Facebook was just running one on their servers rather than you running it on yours.
Ultimately they just briefly used XMPP to not have to implement their own desktop client for their closed proprietary network. It had nothing to do with network reach, unlike Google Talk which did actually bring XMPP to the masses for a while (and then took it away).
Whether you view the question as nonsensical, the most simple example of a riddle, or even an intentional "gotcha" doesn't really matter. The point is that people are asking the LLMs very complex questions where the details are buried even more than this simple example. The answers they get could be completely incorrect, flawed approaches/solutions/designs, or just mildly misguided advice. People are then taking this output and citing it as proof or even objectively correct. I think there are ton of reasons this could be but a particularly destructive reason is that responses are designed to be convincing.
You _could_ say humans output similar answers to questions, but I think that is being intellectually dishonest. Context, experience, observation, objectivity, and actual intelligence is clearly important and not something the LLM has.
It is increasingly frustrating to me why we cannot just use these tools for what they are good for. We have, yet again, allowed big tech to go balls deep into ham-fisting this technology irresponsibly into every facet of our lives the name of capital. Let us not even go into the finances of this shitshow.
Yeah people are always like "these are just trick questions!" as though the correct mode of use for an LLM is quizzing it on things where the answer is already available. Where LLMs have the greatest potential to steer you wrong is when you ask something where the answer is not obvious, the question might be ill-formed, or the user is incorrectly convinced that something should be possible (or easy) when it isn't. Such cases have a lot more in common with these "nonsensical riddles" than they do with any possible frontier benchmark.
This is especially obvious when viewing the reasoning trace for models like Claude, which often spends a lot of time speculating about the user's "hints" and trying to parse out the intent of the user in asking the question. Essentially, the model I use for LLMs these days is to treat them as very good "test takers" which have limited open book access to a large swathe of the internet. They are trying to ace the test by any means necessary and love to take shortcuts to get there that don't require actual "reasoning" (which burns tokens and increases the context window, decreasing accuracy overall). For example, when asked to read a full paper, focusing on the implications for some particular problem, Claude agents will try to cheat by skimming until they get to a section that feels relevant, then searching directly for some words they read in that section. They will do this even if told explicitly that they must read the whole paper. I assume this is because the vast majority of the time, for the kinds of questions that they are trained on, this sort of behavior maximizes their reward function (though I'm sure I'm getting lots of details wrong about the way frontier models are trained, I find it very unlikely that the kinds of prompts that these agents get very closely resemble data found in the wild on the internet pre-LLMs).
I don't share his experience entirely, by even on my desktop built for gaming I can notice the right click menu is delayed in comparison to Windows 10. Even more heinous, before you remove it, the AI button would lazy load causing you to sometimes hit it by accident when you mean to hit something else. God forbid I'm not 80 years old and click my menus with any sort of speed.
Also, if I'm going to have to adjust anything to use an operating system, I might as well use Linux. The only value prop for me to use Windows was gaming, but at this point I'm just completely ripping the band-aid off because it doesn't seem like Microsoft is going in a better direction.
> Even more heinous, before you remove it, the AI button would lazy load causing you to sometimes hit it by accident when you mean to hit something else.
I had a comment a few days ago about my freshly installing, configuring, and A/B testing W10 and W11 on a not-too-old HP laptop, using both HDD and NVMe:
Now last night my partner comes back from the dumpster with a Toshiba laptop that's in great shape, its 20 full years old, came with Windows Vista, had been upgraded to Windows 7, has 1GB of memory, 120GB SATA2 HDD (slower than modern SATA), dual-core 1.73GHz CPU.
Seems to work perfectly and With W7 it's snappier than the 2019 HP booted to old or new Windows 10 or Windows 11 even more. All comparisons at baseline without the internet.
When you try W11 on a HDD it really emphasizes the difference from W10 on the same hardware.
But when this kind of W7 thing just falls in your lap, it really hits you how much better performance could have been available by now from Windows if they really tried.
Or didn't drop the ball every time they were at bat, depending on the general manager ;)
Yes, they could roll their own, but you have no issues with this being necessary? I think the attitude of "just deal with it" is far more negative than someone expressing they are upset with the state of the internet, its controllers, and its abusers.
This is like saying "lets just get rid of all the guns" to solve gun violence and gun crime in the USA. The cat is out of the bag and no one can put it back. We live in a different world now and we have to figure it out.
Cared about anything other than their own upward movement, actively worked towards my professional development, made sure I had actual, not hand wavey, feedback, and made sure my compensation reflected my growing responsibility.
I am aware that all of those things may not be in their power to give, but some combination of that in any org that is somewhat functional would be motivating.
I think you'll find, especially within the tech community, people struggle with purity and semantics. They see that supporting and promoting FOSS is to be okay with its use for war, oppression, or whatever mental gymnastics they need to just not care or promote bad things. They will argue about what "free and open" means and get mixed up in definitions, political alignments, etc.
It is pretty obvious to me, that being blase about whomever using FOSS for adversarial reasons is not very "open" or "free". Somewhere in the thread there is an argument about the paradox of intolerance and I don't really care to argue with people on the internet about it because it is hard to assume the debate is in good faith.
My point is this: Throw away all your self described nuance and ask this yourself whether or not you think any malicious, war-monger, authoritarian, or hyper-capitalist state would permit a free and open source software environment? If the objective of a business, government, or billionaire is power, control, and/or exclusivity then, well, your lofty ideals behind FOSS have completely collapsed.
No I am not. Your response proves my point in regards to getting bogged down in semantics. In a nutshell, my point is that if we do not care or do nothing when it comes to malicious use of FOSS, you very well may lose FOSS or at least the ability to develop in a FOSS environment. It is the paradox of intolerance of a different flavor.
They get to feel like hackerman without understanding any of it. Also, this feels like a security nightmare. I wouldn't self host anything without understanding what you're opening yourself up to.
I had a failing work laptop that had bad battery power and finally just said fine, give me a Mac. The battery and build quality is the only good thing I can say about it. I absolutely hate the OS, despite using MacOS in the past and felt only mildly inconvenienced. It is still amazing to me how unwieldy it is to make keyboard shortcuts, have tiling that isn't embarrassingly bad, and something that is visually consistent. Now I know most people aren't using this tool like I do and Linux has been historically bad at this, but lately, I'm not so sure. KDE and COSMIC seem to handle these cases flawlessly, and even GNOME, which is divisive in the DE discussion seems to get these things right. MacOS/Windows have officially crossed over to being more cumbersome rather than less than your bog standard Linux distro. Have you ever tried to do anything other than adjust volume for your sound settings on Windows 11? It's absurd. You can see the remnants of the Windows 10 attempt at simplifying it with a new flavor of 11 nonsense, and to really do anything meaningful you STILL end up with the old school Control Panel style settings window. A company worth billions couldn't come up with something better for decades. Tahoe is a similar stumble. How does one take these companies serious as a consumer product anymore if you're anything but a casual browser user?
I share your experience. Additionally, I am surprised anyone on this site did not see this progression coming. Between costs, the race to be THE provider, and anyone who has an awareness of how the tech industry has been operating the last 15 years, this move by Anthropic was so laughably predictable that the discourse in this thread is pretty disappointing.
Edit: Seems someone beat me to it with a good reply.
reply