Live data from Hacker News

Drawbacks of P2P and a defense of Signal

changelog.complete.org

201–210 of 215 posts

Re: Drawbacks of P2P and a defense of Signal

#201

Earlier quoted context omitted.

Yep all it took was some usability products and it was easier... See what I'm saying...

In this case centralization is the usability, just like AOL centralized internet access options for users? Maybe you're unfamiliar with AOL in the beginning but it was largely a walled garden that happened to have a web browser.

Centralisation didn't make it easier to use, ask yourself why AOL was in every household

Re: Drawbacks of P2P and a defense of Signal

#202
post #183

Earlier quoted context omitted.

And who knows if it's actually the software they deploy too.

> And who knows if it's actually the software they deploy too. The whole literal point of the protocol is so that you don't have to trust the server , assuming you trust the client and protocol.

That's true for the fundamental part of sending and receiving messages, but at least one ancillary feature (contact discovery) requires trusting the server. They do have a setup with an Intel SGX enclave for remote attestation but I'm not knowledgeable on the limitations of this, although I understand that there are some.

https://signal.org/blog/private-contact-discovery/

Re: Drawbacks of P2P and a defense of Signal

#203

Can anyone chime in on their experiences hosting a synapse server? I have close to a dozen people on mine (although we don't federate much with the network as a whole, I set it up initially for a couple of groupchats/DMs with friends) and I'm not even close to hitting the limits of the $5/mo Digitalocean server I put it on. Does federating with the greater Matrix ecosystem cost that much extra processing power, that…

I run a non-federated instance on a tiny 2GB RAM ARM SOC. About 4-5 active users. Extremely light (under 80MB RAM usage) as long as you don't federate. Attempted enabling federation once - maxed out all cores, all mem/swap, and basically blew up the entire machine. Had to wipe away the database and start without federation.

Encryption UI is a mess. I have actively tried and failed to set up E2E encryption with at least one other technical user. Recommend disabling encryption unless you want non-technical users to abandon ship immediately.

Re: Drawbacks of P2P and a defense of Signal

#204
post #31
post #22

Element/Matrix/Gitter is a great IRC-replacement and public chat, Mastodon is great as a decentralized Twitter alternative, IPFS is great for decentralized public file sharing. Default-public services like this probably have an easier time being decentralized. It'd actually be nice to see thoughtful integrations between these three IMO. The greatest advantage Signal has is probably doing same-day updates for all clie…

> Mastodon is great as a decentralized Twitter Federated, not decentralized. Decentralized is peer2peer like torrents. Federated is like email where there are little fiefdoms like Gmail, yahoo, or self-hosted.

Federation is decentralized. Peer-to-peer is distributed.

Re: Drawbacks of P2P and a defense of Signal

#205
post #183

Earlier quoted context omitted.

> And who knows if it's actually the software they deploy too. The whole literal point of the protocol is so that you don't have to trust the server , assuming you trust the client and protocol.

That's true for the fundamental part of sending and receiving messages, but at least one ancillary feature (contact discovery) requires trusting the server. They do have a setup with an Intel SGX enclave for remote attestation but I'm not knowledgeable on the limitations of this, although I understand that there are some. https://signal.org/blog/private-contact-discovery/

Good point, I was just considering the messaging part. It'd be useful to be able to opt out of contact discovery, as I'm not sure you can do this on the current client.

Re: Drawbacks of P2P and a defense of Signal

#206
post #5

A bit annoying how servers meant to be run by people themselves are written in awfully slow languages. Python for matrix server is a terrible choice. Even worse for home assistant. If it can't fit on raspberry pi it's useless.

You may want to look at Dendrite which is an alternative implementation that is being developed. It's implemented in Go, and is supposedly a lot lighter on system resources.

https://matrix.org/docs/projects/server/dendrite

Re: Drawbacks of P2P and a defense of Signal

#207
post #185

Earlier quoted context omitted.

Do people not realise that lots of single actors have power over all their communications°? Their ISP, for instance. ° Let's gloss over the fact that all here is hyperbole, unless you use a single mechanism / medium for communicating with everyone.

Of course they do. That's why millions of us have been pushing for years for DNSSEC, TLS (hopefully someday with DANE), Tor onion services, RPKI, etc.. A bad state of affairs is no reason to keep going in the wrong direction.

DANE is a PKI whose roots are controlled by world governments, so pushing for it as an alternative to "single actors" doesn't make much sense.

Re: Drawbacks of P2P and a defense of Signal

#208
post #199

Earlier quoted context omitted.

> Essentially you either use the main server or you are a large org that can pay for a powerful machine to run it. To be perfectly clear, the above was the original statement I was referring to and it is completely false.

That statement isn’t strictly true in the most literal sense, but neither is it a clear falsehood, specially given that it is prefixed with the word ‘essentially’, meaning that it is a generalization. It’s also not going to remain true over time. But it’s not a false characterization in practical terms for most people.

I'm saying it's not true in the general case either. In fact, taking into account other things I've said, my claim is that Synapse can easily run on a home computer or a relatively cheap/weak VPS.

In what sense is it then true that you have to be a large organization and have a powerful machine to be able to run it?

Re: Drawbacks of P2P and a defense of Signal

#209
post #55

A benevolent dictatorship would always be more effective than a democracy, but what happens once it stops being benevolent? Same here. A centralized app (like Signal) is more effective than distributed/decentralized approach. But what would happen if Signal would ever stop being benevolent? Remember the days Google were "do no evil"?

Then you switch.

Same, what happens when your Matrix instance stops being benevolent? You switch.

Re: Drawbacks of P2P and a defense of Signal

#210
post #199

Earlier quoted context omitted.

That statement isn’t strictly true in the most literal sense, but neither is it a clear falsehood, specially given that it is prefixed with the word ‘essentially’, meaning that it is a generalization. It’s also not going to remain true over time. But it’s not a false characterization in practical terms for most people.

I'm saying it's not true in the general case either. In fact, taking into account other things I've said, my claim is that Synapse can easily run on a home computer or a relatively cheap/weak VPS. In what sense is it then true that you have to be a large organization and have a powerful machine to be able to run it?

The fact that you are disputed in saying this by multiple comments on this thread suggests it’s not as simple as just accusing someone of spreading falsehoods.

Here is a quote:

“speaking as the project lead for Matrix.

1. It's true that Synapse can use a lot of RAM.”

Post reply on HN