[HN Gopher] The Unrealized Potential of Federation
___________________________________________________________________
The Unrealized Potential of Federation
Author : zdw
Score : 49 points
Date : 2020-09-21 21:41 UTC (1 hours ago)
(HTM) web link (drewdevault.com)
(TXT) w3m dump (drewdevault.com)
| erpellan wrote:
| I was slightly surprised to get all the way through the article
| without seeing a single mention of NNTP.
| hinkley wrote:
| As a solution or as something that was tried before?
|
| Spam as a term transitioned from, "flooding a communication
| channel with the same message" to "ads sent indiscriminately"
| due to a green card advertisement on NNTP.
|
| What we need along with federation is federated moderation.
| There are some characterizations of content where we can agree
| to a label but not whether the label is bad. You might trust me
| to label gore accurately, I might trust you to label incitement
| to doxing. We might never agree on what constitutes baiting. Or
| the crossover point where a flame war should be filtered. So
| you might ban an individual who I think is "keepin it real" or
| vice versa. But ten of us shouldn't have to do the exact same
| moderation in order to maintain a federation. N log n would be
| great, 3N would probably be tolerable.
| asim wrote:
| Strangely I posed a question to HN just a little earlier about
| the future of sofware -
| https://news.ycombinator.com/item?id=24547534 - and you've
| answered that question as I was expecting. I have been
| formulating similar theories and ideas but it's unclear how
| feasible it is without a collective willing to take it the
| distance.
|
| For me it starts with the basis of a communication protocol for
| services https://github.com/asim/mucp. And then moves onto
| something on a broader scale
| https://gist.github.com/asim/ad6e157294d39c2040a5d51ab3d3216....
| I don't have all the answers yet, but I feel something is coming.
| An answer to our future needs and its not p2p as you say. P2P
| might be useful for trade but not services or long term ownership
| and privacy.
| troquerre wrote:
| Traditionally two of the issues with federation have been sharing
| namespaces and sharing logins. Once you need a single source of
| truth, the federation breaks and you need to rely on a central
| authority. There are some new decentralized protocols
| (Handshake's decentralized naming system is one) that provide a
| solution to these problems so I'm bullish on federation in
| general -- I think a federated system could break into the
| mainstream given the recent geopolitical issues affecting
| centralized services.
| asim wrote:
| Is it easier to just use federated identity that creates
| pointers to the origin account? e.g account exists on server A
| and a new signup to server B creates a link as
| acc://alice@server-a/blabla
| ve55 wrote:
| Federated and decentralized solutions have significant UX issues
| compared to leading products in the area. They're not only harder
| to use and slower, but are much worse at converting and keeping
| users to begin with, since they're not ran like businesses are,
| which take in more funding and try hard to optimize all of these
| user metrics as KPI. It will be a long time before this can
| change I think, users still do not (and won't) care if what
| they're using is federated or not. I say this as someone who
| loves the idea of federated platforms, but we cannot truly
| succeed until we have a way to convince large amounts of (normal)
| users to give them a try.
| ddevault wrote:
| On the other hand, the fact that no one is thinking about
| "optimizing KPIs" is a major reason why people _like_ the up
| and coming federated systems.
| warkdarrior wrote:
| > "This [federation] system also makes it hard for marketing and
| spam to get a foothold"
|
| > "You're certainly familiar with another federation which is not
| based on ActivityPub: email."
|
| Self-contradicting article?
| ilammy wrote:
| You can fight spam by refusing to federate with servers which
| are often spamming. That acts as a deterrent for servers to
| host spam since they'll lose participation. However, this only
| works if there are enough of _other_ servers to federate with
| so that you can actually choose your peers.
|
| The status quo with email is not like that, unfortunately.
| Either you federate with GMail and accept the possible spam, or
| you disrupt significant part of communications of your users by
| refusing to accept email from Google's mail servers, with
| Google not losing basically anything (and having no intrinsic
| reason to fight spam). So it becomes a loss-meh situation
| instead of win-loss.
| Semiapies wrote:
| Heh.
|
| Honestly, I'd like to see more yammering about federation by
| people aware of the early federated systems of the internet and
| some perspective on their uses and failings.
|
| Or, better, less yammering and more realizing. Preferably more
| so than small networks where Nazi Twitter is the biggest node.
| littlestymaar wrote:
| Email is particularly broken regarding spam for at least two
| reasons:
|
| - it's a distributed system and not really a federated one: you
| can set up your own SMTP agent and send mails to anyone without
| prior registration to a federation.
|
| - anyone can impersonate anyone: if I want to send an email
| pretending I'm bill.gates@apple.com there's no built-in
| mechanism to prevents that.
|
| Theses aren't fundamental features of federations, and came
| from a design which just overlooked how many bad actors email
| would attract when it became mainstream.
| ddevault wrote:
| Email is a bit different from the new federations we're seeing
| today: there's less weaker guarantees of identity, and there's
| less emphasis on choosing what other instances you federate
| with. In the example of Mastodon, the cost of setting up a new
| server is also too high to make spam effective, since you need
| a persistent, reachable account in order to interact with other
| instances on behalf of it. And the social factor counts, too,
| newer federations like this tend to be opposed to corporate
| influence in general and this makes it hard for it to get the
| initial foothold.
|
| Email has a lot of spam, but a federated system which keeps
| this in mind from the outset needn't suffer the same fate.
|
| EDIT: Incorporating the replies to my comment, I also want to
| mention that the lack of large, centralized players like gmail
| also prevents spam. If there are a lot of intances, and they're
| all small, then cutting one off for bad behavior is a lot less
| difficult. Instance admins can be held accountable for the
| behavior of their users.
| riffic wrote:
| note that you _do not need_ a full-blown Mastodon server to
| participate in the AP ecosystem. There are very little
| barriers to entry here - one is being able to process dynamic
| scripts and the other is a registered domain name.
|
| I think the lowest cost solution to this would be a WordPress
| site on a shared host somewhere, with this plugin installed:
|
| https://wordpress.org/plugins/activitypub/
| upofadown wrote:
| Email is one to one mostly. So reputation should really also
| be based on individual entities. The normal way to make that
| possible is to have email clients sign messages by default
| with the senders identity of choice.
|
| The technology exists in the current system. We just need to
| activate it. This isn't a technical problem.
| Jasper_ wrote:
| That would imply that centralized solutions would have even
| less spam, since they have much stronger guarantees of
| identity. And yet sites like Twitter has plenty of spam. Even
| Hacker News gets spam! And this is not a popular place in the
| grand scheme of things.
|
| How is distributed identity solved in the long term? Nobody
| would run a server to communicate spam; that's rarely how
| email spam is sent these days. Most of the spam I get is from
| accounts hosted on GMail and Yahoo!. So the weakest link is a
| single server that's community-respected, but doesn't have
| adequate spam protection. Federation makes that _much_
| harder, because the whole point is that every server is run
| separately.
|
| Mastodon gets no spam today because it's smaller than Hacker
| News. There's relatively nobody on there.
| kzrdude wrote:
| I'm worried how mastodon is going to scale (in terms of
| moderation) - maybe some instance can experient with crowd
| sourced moderation.
| riffic wrote:
| > How is distributed identity solved in the long term?
|
| Same way email handles it -- through the use of the domain
| name system. You'd expect that an a federated server with
| the address "https://mastodon.mit.edu/" would consist of
| those with a tie to the Massachusetts Institute of
| Technology (faculty or students). Your identity is provided
| by the system you are a member of and handled by your
| specific system administrator.
| jethro_tell wrote:
| Email suffers from one of it's early defining operational
| principals which was to be backwards compatible forever. In
| this way, operators still have to decide if they should
| accept federation with weakly identified entities or close
| the door and federate only with high reputation sources. To
| add to that, some forms of it's adoption are difficult to
| work around with normal methods. Mailing lists are tough,
| since they spoof the senders address and break dkim. And once
| again, the decision between allowing the old way, and just
| blocking anyone who doesn't have proper dkim.
|
| We see the complaints about this every time someone posts
| about an email server. It's hard to get the big guys to play
| ball. On the other hand, every time email comes up, people
| complain that there's too much spam.
| CKN23-ARIN wrote:
| People often forget that "the Internet" itself (i.e. IP, BGP)
| operates on a federated basis.
___________________________________________________________________
(page generated 2020-09-21 23:00 UTC)