Federation in social networks, why it's never going to work.

Posted Dec 14, 2017 20:31 UTC (Thu) by zarrro (guest, #54749)
In reply to: Federation in social networks, why it's never going to work. by jejb
Parent article: Federation in social networks

> if they could get away with it they'd kill SMTP federation as well.

I think this is slowly happening. The barrier to having your own email server is getting higher and higher.

I mean yes, you can buy a domain, install a server, but the chances are very high that your emails will end up in spam in at least Gmail.

Federation in social networks, why it's never going to work.

Posted Dec 14, 2017 21:33 UTC (Thu) by jcrawfordor (guest, #114167) [Link] (2 responses)

I sadly agree with you here, when people ask me about running their own mail I tell them the easiest way to run a mail server is to pay someone else to do it!

I think this illuminates one of the big problems with federated systems... email has become the maintenance nightmare it is today for a few reasons, many of which are direct results of federation: ad-hoc methods of combating spam, and compatibility issues that tend to prevent making any changes to the protocol without nasty hacks around old implementations.

Newer systems seem to be a little less sensitive on the latter issue because most newer federation systems are using markup languages and other kinds of structured data that will encourage older implementations to just ignore content they don't understand.

As for spam, the OStatus/ActivitySub system does have a bit of a trick there. Because instances generally only start subscribing to other instances when one of their users follow a user on that other instance, new instances that appear on the network just to produce spam will tend not to be subscribed to by anyone. It's still possible for spammers to do things like pushing direct messages, but the attack surface is a little smaller than it might seem. Doing something like rejecting all pushes from an instance that no one on the current instance subscribes to are reasonably low cost measures that might seriously knock down spam potential.

Federation in social networks, why it's never going to work.

Posted Dec 15, 2017 10:51 UTC (Fri) by NAR (subscriber, #1313) [Link] (1 responses)

"As for spam, the OStatus/ActivitySub system does have a bit of a trick there. Because instances generally only start subscribing to other instances when one of their users follow a user on that other instance, new instances that appear on the network just to produce spam will tend not to be subscribed to by anyone."

My only knowledge about microblogging is that Twitter has a really high profile user, but don't these microblogs get comments or answers? In that case the spammers can send their spam in comments and answers...

Federation in social networks, why it's inevitable

Posted Dec 21, 2017 3:55 UTC (Thu) by Garak (guest, #99377) [Link]

"don't these microblogs get comments or answers? In that case the spammers can send their spam in comments and answers"

At the end of the day it's web-of-trust moderation. White/pass-lists and Black/block-lists and experiments with various graylisting techniques in between. Sources of spam get moderated/untrusted, perhaps fractionally and incrementally. Spam was never a problem for people who used whitelists of contacts that go to their main inbox with everything else going to a check-once-in-a-blue-moon folder if not /dev/null.

Also, 'microblogging' is stupid. Add a feature to make a default display have a message size limit, with a one-click away from 'more' and you can support both users with not much to say, and lots to say.

Finally, I consider the widespread prohibition of home/mobile servers by ISPs to be the real impediment. If that barrier were removed, development would happen exponentially faster. As in 5 years ago we'd have had stuff way better than we will in reality 5 years from now. The development that goes on with the w3c's blessing is political distraction from the field of opportunities to experiment everyone should have with their ordinary home/mobile internet service. The internet is and should be more than just http/s. It should be a place where people are free to develop and experiment with new protocols without having to negotiate permission from their ISP/gatekeeper.