moxie seems to believe there are only two options for how services can be hosted: self-hosting or commercial hosting.

completely ignoring community hosting.

that reinforces my feel that community hosting is one of the biggest advantages of fedi.

(he is absolutely right that is crap, of course)

@rysiek even though I see (and agree) with your point, I think it might be more subtle than that:

- individuals or communities can have their own fedi instances
- they then choose how they have it: either they self-host it, or they use commercial hosting

@vanecx oh absolutely! but a community instance hosted on is a completely different beast than a community Facebook group. and that's my point, which moxie seems to miss altogether.

@rysiek @vanecx I do think his final two points are decent, though, especially the one about designing things so that people can have assurance/'distributed trust' without having to run their own servers (e.g. "download entire blockchain" in this sphere).

@meejah @vanecx yeah, absolutely. And, for example, account portability on Matrix is kind of sort of going this way, I guess?

@vanecx @rysiek I think that the problem is that many individuals don't care about the fungibility of their technology.

Commercial hosters will always have an incentive to make their offering non-fungible (i.e. offer features that aren't replicated by their competitors). Thus creating lock-in, even if that lock-in preserves interoperability at first.

@michiel @vanecx I think the problem is that many individuals don't care about the non-toxicity if their food.

Commercial food providers will always have an incentive to make their offering cheaper (thus likely toxic).

...Apologies for this little gedankenexperiment. My point is: we care about things we were taught to care about. People don't understand why privacy and standards *matter*. Education plays a huge part.

@rysiek Agree. Unfortunately however, a lot of emphasis is put on self-hosting rather than community hosting these days and "run your own stuff, it's not that difficult" is am advice much more common than "support a hosting community". Consequently, there's a lack of structure and culture of building and fostering communities, at least from what I see. Not even talking about communities doing more than "just" running software.

@z428 agreed, although there is plenty of great communities around. the ones that seem the most successful are the ones that started off as actual communities and then decided "hey how about we run a fedi instance for our members".

we need more of that!

@rysiek it is not community hosting that I am most concerned about, when it comes to fedi.

It is grassroots-community-based open standard specs and the evolution thereof.

Here Moxie has a point. After AS/AP became W3C recommendations, there's been very little in terms of follow-up. Interop mostly involves looking in foreign codebases and trying to find a good expertised dev to explain the undocumented details.

And with every new app + custom choices, the interop landscape becomes more complex

@humanetech sure. this is exactly why XMPP is such a mess. there is now some movement to standardize certain collections of XEPs, which would make all the sense. I am hoping fedi doesn't have to go through a similar cycle to get this right though.

Although it's a fair point that a protocol tends to ossify around a lowest-common denominator system that is stagnant, this doesn't prevent individual clients from advancing.

Also, clients can support new protocols as they arise.

The most important point is still that this avoids getting trapped by a monopoly. ISTM that any technical agility that closed silos might provide has to be weighed against that social benefit.

@humanetech @rysiek

I ran out of characters to include examples, which illustrates the criticism, I suppose. But there are Fediverse clients with longer limits (I assume, because I see longer posts).

And I was going to mention both the wide range of email clients from minimal to sophisticated management tools.

And on the point of stacking protocols, email clients that also read news, rss feeds, etc.

(Supporting that protocols are not necessarily a serious impediment to innovation)

@TerryHancock @humanetech @rysiek It's absolutely easier to change protocols in a monopoly; there's no need to come to consensus with any other groups. Consensus is hard.
Still, further to Terry's point, I do agree that open protocols with a wider ecosystem are _better_

@meejah @TerryHancock @humanetech @rysiek Right, evolving a protocol is a social problem, and writing code was traditionally easier than to break into the ivory tower of standardisation processes. I think it's (in part) a matter of making standardisation bodies + processes more accessible to a bigger audience. Once the friction between protocol design and implementation decreases, I think federated systems have a chance.

The early success of the internet and the rfc system was, in my opinion, down to making protocols and standards a low barrier and fast turnaround process. It enabled IETF to run circles around ITU and ISO.
@meejah @TerryHancock @humanetech @rysiek

@maswan @eaon @meejah @TerryHancock @rysiek

Yes, you are all right. And a grassroots community slowly evolving the ecosystem with their app-specific additions is where we are now.

But if there's no coordination to the the substrate on which this is built, then I fear a Web2 trend of convergence around the most popular apps will be the result.

We see that with Mastodon and seeing other apps with "I just copy/paste the Mastodon API" and "Add in these masto msg props to interoperate".

@humanetech @maswan @eaon @meejah @TerryHancock @rysiek This is something I'm very concerned with. It's complicated to do this "right", I suppose, but the last years have shown me quite clearly that a) someone has to start trying, and b) a lot of folk want it to happen. Moxie would be in a good position to start, but instead claims it's too hard. I find that sad.

A few people do try.

@jens @maswan @eaon @meejah @TerryHancock @rysiek

Just read response to Moxie by André Staltz of #SSB / #Manyverse:

Think when it comes to #Fediverse then the #SocialHub community made a good start, but to take it further there's need for dedicated people to keep the innovation processes going and towards #interoperability and #interconnectivity

Forging something that rises beyond apps and into #SocialFabric formation. Something where everyone can find their place.

@humanetech @maswan @eaon @meejah @TerryHancock @rysiek My background of course colours my perception on this, but I'm convinced a lot of folk wouldn't mind running their own servers if they just didn't know (aka P2P), and a lot of other people wouldn't mind running servers for communities that just provide storage/relay capabilities, with no hosting of software in any web sense.


That’s probably the largest difference between federated services (such as Fediverse) and full P2P (such as SSB). Even running a SSB node is quite challenging: IPv4 NAT, extremely fast growth of disk space usage, high network usage, high resource usage. I have been using SSB, including running a pub and a few nodes, for 1-2 years and just gave up. The primary lesson drawn from this can be summarised in one sentence: servers belong to data centers. An average user’s home network is not designed to be a data center, from both networking and security point of view. When I ended up shutting down the SSB node to be able to reasonably use SSH, I understood that’s probably not the best design from purely practical point of view.

Right now I’m running Matrix, Pleroma and a dozen of other federated services - but all of them live in a data center, while I connect to them using mobile & desktop clients even from a very slow mobile network.

@humanetech @maswan @eaon @meejah @TerryHancock @rysiek

One of the things I like to bring in from our computer club @acc is providing heavy infrastructure for community projects. Especially things that are easy to run but uses lots of hardware or bandwidth, like our free software and media mirror and other things, like a few IRC networks, Open Streetmap, mastodon servers, etc.

There are lots of community members, and they can bring different types of resources.
@eaon @TerryHancock @humanetech @meejah @rysiek @jens

@kravietz @eaon @maswan @TerryHancock @humanetech @meejah @rysiek I'm not saying they're not problems (they are), but they're problems of resource management, not architecture. We managed to shuffle around video data efficiently enough in 2006 that these issues weren't noticed often. But I do of course agree that such concerns need to be addressed appropriately.

@jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek Remotely using a computer (maintaining a server), dealing with DNS, and the monthly cost of it to a degree is definitely a high barrier to entry for many people.

ISP-provided personal URLs + a "self-hosting" router that had simple/secure install setups for server apps would probably open up home networking a lot.

@edd agree with this, except i wouldnt want the ISP involved beyond the very basic role of providing a connection to the internet. In the US all ISPs are basically profiteering turds.

@jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek

I don't disagree, but I'm thinking something akin to ISP provided email, e.g., that points to your router without needing to update your DNS setting as your lease expires or pay for a static IP. Just simplifies it for non-tech folks and lets them dip a toe in first.

@jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek

Maintaining a server is not for everyone, but there are projects like @yunohost and @freedomboxfndn which makes it nearly as easy as installing a distro for intermediate-level users.

Somewhere I read that their next phase is to create hardware for sale that just plug and play and it works, just like WiFi routers.

@jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek

Example: I have one VPS running that provides (now via yunohost) mail, nextcloud and some other services to my family. This makes 1 server for 23 persons. 1 out of 23 person need some IT skills.
The architecture could scale up to 200 persons before we would need extra stuff like load balancers etc.

@edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek

@ashwinvis @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

I see the 'community hosting' model is that you get this service from a local or virtual trusted community, whether commercial or not, on whatever business or service model your community favors, rather than from a global corporation that sees you as nothing but raw material, gives you no choices, and makes money off of you how they see fit. That's a valuable difference.

@caliandroid @ashwinvis @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

I'd estimate the cost is something like $100/month in services, plus one half-time job (maybe just quarter time) to support a 25 users, up to a bit more for 200. Maybe $1500/month total, or 25@$60/mo to 200@$7.50/mo -- paid, volunteer, or subsidized.

Plenty of ways to support that level of cost, whether passing the hat, taxation, charging fees, advertising, or business 'goodwill'.


I pay ~15€ per month all-in.
After having set up everything and with heavy automation, I spend maybe one hour per month for the private cloud service - no moderation needed.

I had once a huge bulletin board (~ 20 years ago) with 10k active users. I've spent at least one hour per day and had a team of 30 moderators. Everyone spent at least one hour per day on the board.

@ashwinvis @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

I was running it on a Raspberry Pi 3B+ at my home. Thankfulky our ISP does not block any ports. The monetary costs were therefore negligible. Performance and reliability left more to be desired though.

@TerryHancock @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

My issue with this line of thinking is that IT skills aren't uniformly dispersed geographically, and not all of them want to do sys admin on the side. To really spread out of a localized technical bubble there needs to be low cost, turn-key options that e.g. a casual knitting circle can set up without needing a friend or family member to maintain.

@ashwinvis @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @TerryHancock @rysiek

@edd Luckily, the solution to this could easily be the same. One thing p2p systems tend to do, precisely because they were usually packed within consumer applications, is set themselves up easily. @caliandroid @ashwinvis @yunohost @freedomboxfndn @humanetech @maswan @eaon @meejah @TerryHancock @rysiek


In terms of labor/overhead costs, it's similar to providing free wi-fi. And probably supportable for similar reasons.

And, by far, the biggest cost is moderation & administration time, rather than hardware/service costs. Similar to the demands of running a private forum, but more valuable, because it's more connected.

@ashwinvis @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

You're right but I would say the labor costs are my hobby that I donate with pleisure (and maybe some kind of selfcare because I just love to setup tec, corrupt it and reconfigure it even better than before ;-)

@ashwinvis @edd @yunohost @freedomboxfndn @jens @humanetech @maswan @eaon @meejah @rysiek

@jens @humanetech @maswan @eaon @meejah @rysiek

While it's unlikely everyone is going to want to run servers, even if only 1% DO, that's still going to be many thousands of times more choices than you get with a single-silo platform like Twitter or Facebook serving the whole world, and still escapes the monopoly supplier issue.

So it's more about making it attractive to a 1% tech-savvy fraction of the community. I think that's probably an achievable goal.

@TerryHancock @jens @humanetech @maswan @eaon @meejah attractive *and* sustainable in the very-long-term (think: decades). These are both pretty difficult (doesn't mean we should not try!).

IMHO, a major goal should be marketing the value of hosting a Fedi server, rather than simply making it easier and assuming the benefits are self-evident.

I self-host several apps I need on my project site (Wordpress, Resource Space, Subversion, Trac, MediaWiki). But why, other than 'selfless charity', would I run Mastodon etc, rather than free riding? What's my self-interested motivation? Goodwill/promotion? Control? Ad revenue? Integration?

@jens @humanetech @maswan @eaon @meejah

@TerryHancock @jens @humanetech @maswan @eaon @meejah @rysiek
Definitly. But at the Moment it is too cumbersome. I'm running a matrix Server, but Mastodon is too heavy

@humanetech @jens @maswan @eaon @meejah @TerryHancock @rysiek

The response by @staltz is correct, only a minority will contribute to a community but it's that active minority which makes communities possible.

The YouTube example is perfect: only a minority upload videos yet those are the videos the majority watches.

Moxie doesn't seem to realise (or perhaps doesn't want to admit) that human beings depend on an active community-spirited minority. Society isn't possible without such people.

@FediFollows Maybe core issue here would be finding a meaningful definition of what a "community" is and how this community is organised. For quite some aspects of day-to-day life in quite some parts of the world, I'd argue that communities don't matter much whereas people depend on "products" provided by anonymous individuals or large structures. That's exactly why GAFAM win over community based infrastructure so far.

@humanetech @jens @maswan @eaon @meejah @TerryHancock @rysiek @staltz

@z428 Well, IMHO the first step is to distinguish between real world communities and digital ones. The client/server model conflates the two. It defines community as everyone on the same server (and may allow for subcommunities). In a P2P model, a community is ...

@FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

... closer to the people you've given full trust in your GPG keyring, which is probably still a combination of several logical groups because GPG is rather binary here.

Ideally I'd be able to just run a storage server for everyone I've tagged in a comparable way.

@z428 @FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz


And there doesn't have to be exclusivity here. One could stripe data across some or all instances one has storage capacity on or some such, allowing for redundancy/failover, as well as commercial providers.

@FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

@jens Hmmm, agreed, but that wasn't really what I meant. 🙂 I was more or less thinking along the lines of experience "we" as a society still have with providing "community" based services. Maybe defining "community" as "city", "neighborhood" or "village" (who provides their inhabitants with water, heat, street cleaning, maybe power and health)?

@FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

@z428 Fair enough. But it sort of ties in. In my extended family, I'm roughly the only techhead. If I define the community I want to contribute to as the rest of them, it doesn't really need much more than that, extended to other families, to have huge coverage. @FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

@jens Yes. But: Not all families have tech-heads. And even if so, tech-heads have conflicting ideas. Just imagine your families tech head being a Matrix supporter but your chess club that you want to interact with had decided to go XMPP. Here, it already mostly breaks (or gets quite complex and fragile, talking bridges at all).

@FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

@z428 Oh, sure. I'm not claiming that this is a complete solution to the world's problems. It's a direction. And a way of letting the real world set or remove digital community boundaries instead of following some central service's design, good or bad.

@FediFollows @humanetech @maswan @eaon @meejah @TerryHancock @rysiek @staltz

> Consensus is hard.

if only we had a solution for the problem of distributed consensus... wait!
Sign in to participate in the conversation
Mastodon for Tech Folks

This Mastodon instance is for people interested in technology. Discussions aren't limited to technology, because tech folks shouldn't be limited to technology either!