Hacker News Reader

Comment by swalladge | Hacker News Reader

It's great to see 3/4 of the options are open source! Whatever happens, I really hope the community get behind the open source options and don't let more things get eaten up by commercial silos cough slack cough.

I'm partial towards Matrix/Riot.im - the progress made on those projects is awesome and they really have a highly usable product, with bonuses such as e2e encryption and federation.

View on HN

> they really have a highly usable product

Setting up a Matrix server is a lot harder than you might think, especially when you start talking about federation and identity management.

The mxisd[1] project has recently disbanded due to what I believe to be philosophical differences with the Matrix maintainers vision of identity management.

I like a lot of things about Riot/Synapse, but I would suggest you try setting up your own Matrix server if you haven't yet. It's not what I would call highly usable.

Mind you, I have only tried setting up the reference implementation (Synapse/Riot). I would be interested in seeing a write-up/comparison done by Mozilla as part of these trials.

[1] https://github.com/kamax-matrix/mxisd

I've actually had the opposite experience - it was easy enough to download/set up, and then it was just a matter of getting the letsencrypt daemon on the box to manage certs.

That said, I went in with pretty low expectations since it's a reference implementation - I think one of the great advantages of the open-source protocol is that anyone can design/build their own implementations with better UX on the sysadmin side. There are already projects in the works written in golang/rust, and I'm sure if the protocol takes off other languages will follow.

I do agree that the identity server piece is weird. I didn't set one of these up, so I haven't really looked into it (I don't have any 3rd party IDs connected to my matrix username, which from my cursory research is what the identity servers are for) but the philosophy behind them seems to go against the federation narrative the rest of the protocol is designed around. If anyone has done more research into this part of the protocol I'd love to be corrected on this point.

> is that anyone can design/build their own implementations with better UX

That's a common fallacy. Nobody is going to do for free this crucial work for a protocol that is practically unused. Teams should focus more on a strong implementation so the protocol gains ground

It was pretty straight forward to install/set up? There are some requirements to be able to federate, namely having a domain and associated certs for encryption, but otherwise nothing stood out as overtly obtuse/difficult.

You have to either make a venv for a binary or just install a package, then reverse proxy 2 ports from your http daemon, and finally also hook things up to your Postgres database. After all that it's just a matter of configuring your homeserver's config (which is large, but you can easily get support via the community for any poorly documented options)

The documentation could definitely be improved, though.

If you don't mind Docker, I success this ansible playbook: https://github.com/spantaleev/matrix-docker-ansible-deploy

Great documentation and sane defaults. I have only 6 lines of custom config.

there’s a successor fork to that identity server over at https://github.com/ma1uta/ma1sd which is being well maintained :)

sorry that setting up a homeserver was hard; we’ve been doing a lot of work recently to improve this (eg https://github.com/matrix-org/synapse-config-generator/tree/... is a graphical installer which should be released very shortly).

This looks good. Will this become part of matrix in the future (also, does it work /stable?)?

the config generator will ship in synapse by default in the near future. its first release is due tomorrow.

ma1sd is considered stable and works afaik, and is already part of the matrix ecosystem.

Thanks, I was not aware of ma1sd. I saw Gridepo but it seems to solve different problems.

I felt as though getting Matrix 95% up and running was relatively straight-forward, but I spent far too long trying to configure self-registration, email verification and domain whitelisting through mxisd that I gave up. These don't seem like uncommon scenarios for organisations wishing to run their own Matrix instance.

In contrast, a Mattermost instance that supported self-registration and domain whitelisting was up and running in almost no time.

The config generator will be a welcome addition. I will probably revisit Matrix some time in the future.

This is really frustrating. I was hoping I could replace ZNC with Matrix plus my own IRC bridge. Yes I know about Matrix's bridge.

You certainly can for individual use. My experience was more about setting up an enterprise-grade server with identity management.

Yeah that sucks. I wanted to integrate matrix to our community but no, it would be a bad decision at this time

Interesting. Yeah I've set up a synapse server and maintained a small instance for a while in the past. I didn't find it overly difficult, and I hear they've made usability improvements since then. I didn't try anything advanced with identity management though (and don't know much about it apart from the fact that it's not decentralized yet).

I actually switched my whole family to a private Matrix server after the Hangouts retirement announcement. It's got some rough edges (especially identity management) but it works well enough overall and it was the best option of all the chat apps I looked at.

> I really hope the community get behind the open source options and don't let more things get eaten up by commercial silos cough slack cough.

Don't forget Slack also bans people who they think are from sanctioned countries:

https://www.theverge.com/2018/12/20/18150129/slack-iran-deac...

I would like to see a neutral platform be chosen to be honest.

Yeah wow that too. :O

> I would like to see a neutral platform be chosen to be honest.

Exactly, and if one of the FLOSS options are chosen, Mozilla could self-host and not need to rely on any third party.

There's a fair bit of weight getting behind Matrix too, with the french government starting to use it for internal communication.

Yea! My next prospective smartphone is doubling down on Matrix: https://matrix.org/blog/2017/08/24/the-librem-5-from-purism-...

Until Matrix/Riot resolves their problem of having a centralized identity server[0] I wouldn't recommend it.

[0]:https://gist.github.com/maxidorius/5736fd09c9194b7a6dc03b6b8...

Yeah that's true, if you wanted a private server a la mattermost/riot/slack, it wouldn't be ideal. However, I feel it's a decent tradeoff at present for what it's aiming for.

Matrix does not require a centralized identity server (which exists only for optionally discovering users on Matrix based on their email/phone numbers).

The linked gist is loudly complaining (amongst other things) that Riot's default config points at a centralised one by default.

We've already fixed this (https://github.com/vector-im/riot-web/issues/10553) by simply removing the concept of default identity servers entirely and instead prompting the user to select on demand whichever one they want to use, if they actually try to look up a user based on email/phone. It should be released in Riot 1.4 (on Web, and equivalent ones on mobile etc) in the coming weeks.

You can see our response to that original gist at https://matrix.org/blog/2019/06/30/tightening-up-privacy-in-..., and you can track progress over at https://vector-im.github.io/feature-dashboard/#/plan?label=p... (which has a cache from yesterday published at https://matrix.org/~matthew/privacy-sprint.html).

Reply on news.ycombinator.com