Hacker News Reader

Comment by Taek | Hacker News Reader

Signal is hosted over centralized servers, I don't think it's that much of a break from the status quo in terms of control.

That's not to say it's not the best out there. As a company, Signal is a million times better and more deserving of our trust than almost any other option on the market, but if we want to break out into open communications, the Signal protocol needs better decentralization.

View on HN

In theory I like the idea of decentralization but in practice I haven't seen it work out well. For example email didn't become prolific until large centralized players became the dominant force. I see federation/decentralization being better as a late feature in software. Through centralization you can move a lot faster, which in the case of Signal is extremely useful. They wouldn't be able to provide the product they have with decentralization. That's why Moxie has been against decentralization. Maybe I'm wrong, but I don't see chat apps as having an unbounded number of features. As I see it once Signal moves into more of maintenance mode then we can talk about federation, by allowing outside servers being able to communicate to the main Signal server. But for now it seems better to focus on getting people to the platform and if we're talking about it from a privacy and security point of view, this is the higher priority. Decentralization doesn't provide much more security while Signal is still trustworthy. There's bigger fish to fry right now.

For what it's worth I think you are wrong. Decentralization is a complicated technology that needs a lot longer to mature than I think most people realize. But once it is mature - and we're getting close, imo just a few years left - it'll move way faster and more efficiently than the centralized world ever did.

In many ways it feels like decentralization is the same as discovering the internal combustion engine before modern alloys existed. We have to invent a lot of extra stuff to make decentralization practical, and we will, but the vision definitely started many steps ahead of the technological progress.

> For what it's worth I think you are wrong.

I'm happy for disagreeing opinions.

> But once it is mature

You may be right, and if it turns out that you are I'll happily change my opinion. But it seems like you agree that currently centralization is the way to go for an app like Signal. Honestly I'd love to see Signal decentralized, but I don't think it is the right move right now or in the near future. But given what you said I still don't think it means it is the right time to call for decentralization. Signal doesn't have to be the leader in advancing decentralization technology. They are already leading in secure and private communication and I don't think such a small group can be experts in too many things. They got their niche and I'm happy someone is doing it.

I'm curious though, why do you think decentralization will move faster in the future? What advantage do you see in that? I've always seen decentralization as lagging in updates because it is more difficult to get users to update and software can't exist in isolation (software rots).

Open data models allow for faster innovation and stronger collaboration. That's basically the meat of it.

Right now the cost outweighs the benefit but the cost is decreasing and the benefit remains.

Sure, you can potentially innovate faster, but unless you've got a good way of enforcing your changes on the whole network, one of two things happen. Either you either end up forking the network into multiple islands or building some crazy complex "negotiation" mechanism where two random clients need to agree on what features they support before they communicate.

We saw incompatible islands appear in decentralised systems before (this tends to not be good for users), and we've also seen how any implementation of "protocol/feature negotiation" is basically a disaster for security. I don't currently see how any truly decentralised system is lowering either of these costs.

WhatsApp never used to be e2e for everyone, but they were able to make it universal relatively quickly because it's centralised. All of the poster child decentralised protocols (SMTP, XMPP etc.) are still not e2e for everyone despite decades of effort.

I think in practice it won't be so much about agreeing on a protocol but rather new apps just parasiting off of the data models of the most popular and successful apps.

The thing about decentralization is that it forces the data to be available. So for example Facebook can't stop an alternative app from having full access to Facebook groups even if they want to.

So the big successful apps most likely end up deciding the data model and feature set, and everyone else just borrows that, and if they extend or add features those features won't appear on every interface. Which is okay as long as the common core (determined by the big app) is all there.

XMPP was mature enough, Matrix is mature enough.

Technology is ready, what's missing is the goodwill of the biggest actors.

XMPP is mature. I think that the downfall is the complexity coming from having almost everything defined as an optional feature on a protocol level, and at a very fine granularity at that. So an implementation has to cope with situations where any other party it wants to communicate with may have an oddball collection of supported features. Plus, some server operators added their own peculiarities on top, like wanting to federate with some servers and not others and so on. The user facing result was a user experience with seemingly random inconsistencies. For example, user A can send a message to user B while user B is offline, but not to user C because C's server doesn't support offline message storage.

> For example email didn't become prolific until large centralized players became the dominant force.

What? Email was massively popular before gmail even got into the game. It was popular when Microsoft bought hotmail. It was so popular I remember a dial-up ISP (Juno) offering a free email-only dial-up account.

Email was one of the only easy to convey selling points for early Internet.

Online shopping was a shitty Wild West and most web pages consisted of some basic info and a visitor counter. But being able to send a “love letter” across the country in seconds was amazing!

> It was popular when Microsoft bought hotmail

I'm sorry, but is Hotmail not a large centralized player?

I think the GP was saying that email was already popular before Hotmail entered the picture. I agree with the claim that it didn't need centralized players to become popular.

I don't think this is accurate though. I wouldn't call the internet popular when it was just nerds on it. At least not in the context of what we're discussing (communication clients). That's not popular until the masses join. It is clear that the masses did not join until there were large centralized services like AOL, Netscape, Yahoo, and Hotmail that provided email services. I am fundamentally disagreeing with the premise that email became popular before large centralized players provided the services.

And you’re wrong. Email was the main selling point to get non-nerds online in the late 90s. It was how we got my mother in-law online, my uncles, and my non-techie friends.

Lol. Email was the internet for most people when you were in diapers. It attracted federal regulation as early as 1986.

Decentralization is hard to build a business around.

Reply on news.ycombinator.com