I don't think this is a fair characterisation of the article. She argues that email is an example of something that is failing as a decentralised application: Google's power over email is too strong. This is very different from claiming that it isn't an example of a decentralised application.
Email and the web are decentralised applications. But they are not the email and the web of the 1990s, when their decentralisation was the killer feature that saw them beat out centralised networks. To try and live on a non-Google/non-Facebook web is like using Linux on the desktop in 2000: it's certainly possible, but it's a bit of a hairshirt.
Spam filtering is a a very difficult thing to do well and Google's implementation was very good, very advanced, and trained on huge amounts of data. Something a normal person could never do on their own server and those who could would be spending a lot of time training the blocker which those using Google never even needed to think of.
Public organizational emails in orgs I've been a part of (e.g. [email protected], [email protected]) get a little more spam, largely sales pitches from what appears to be actual humans at shady SEO companies. Maybe 2 a week.
I get way more "legitimate" corporate spam than actual unprompted/cold spam. The most egregious example in recent memory was when I made an online purchase at Bed, Bath, & Beyond, and made sure the "sign me up for the email list" checkbox was not selected. They sent me no fewer than four emails asking if I would like to sign up for their email list, including one titled, "Thanks for signing up for our email list!" (inside: "Just take this one last step to confirm your subscription <signup link>").
In conclusion, I perceive unwanted email to be largely a self-inflicted problem at present. For example, someone I know likes to complain about the quantity of email she receives, but also refuses to unsubscribe, because sometimes she sees something interesting in one of the many promotional lists she's on. I also don't understand people who say they're not bothered by advertising, but to each her own.
I think this really sums up Google's M.O. in general.
Everyone else had to up their game due to Gmail. On all fronts.
meanwhile tons of spam (crypto related) is marked as spam, without that it'd be pain enter my mailbox
Obviously a personal anecdote, but I signed up for Gmail close to when it first came out, and I don't think I've ever had a legitimate piece of mail marked as spam.
I remember people clamoring to sign up for Gmail because it was the new thing, it had the counter that kept going up with "a lot" of storage space, etc. Folks were even paying others for a Gmail invite.
This, definitely. Take a look at gmail spam folder, it's always full of false positives.
I think that’s “especially as”. Other mail providers are competition, after all.
A slightly more problematic is targeted phishing attacks, where the attackers try to put at least some work-relevant information “. Some of those attempts are not so trivial. I do receive them few times per year, while with GMail I do not remember getting one in at least a year. But on other hand I am not so sure that Google will defend against those if we have used GMail at work.
I wish this meme could die. gmail spam filtering isn't particularly good, it never has been. They crank up the false positive rate high enough that the spam in your inbox is about comparable to other solutions, but at the expense of tons of false positives.
I've had a gmail account since the launch. Been there got the tshirt (literally.. have a tshirt of the launch when they were bragging about 1GB storage). I've also been managing email servers and mailing lists since the very early 90s to today.
Notice every site that interacts with email has the ubiquitous warning abot "check your spam folder"? Not solely, but largely a legacy of so many users on gmail getting used to so much legitimate email going to spam. it doesn't have to be that way.
You can easily to an order of magnitude better on fewer false positives while still getting less spam overall in your inbox, by running your own email servers.
I haven't had a false positive in ... certainly many years on my own email on my servers. And a bit less spam getting through than on the gmail account.
> Getting filtered by Google because of the massive amount of spam they are blocking.
gmail has a big false positive problem, yes, but it isn't any worse for you as a sender if you run your own email servers. I experience more false positives emailing from work account to work account (both sides hosted by gmail) than from personal account (hosted by me) to gmail.
> Spam filtering is a a very difficult thing to do well
Spam filtering is actually not hard anymore. I've been running email servers and mailing lists from before spam existed, through its rise and later fall. There was a time, ca.2000, when it was hard. Because nearly every legitimate email server was misconfigured, so one had to allow it all through, but most was spam. So it took a lot of client side filtering.
These days, very easy. Legitimate senders are well configured, so just reject all misconfigured clients at the SMTP connection. There, that's 95%+ of spam blocked. The few remaining items are easily filtered by whichever bayesian filter you like. I'm using spamprobe. Done. No spam. No false positives.
I'm sure most normal people could install SpamAssassin and configure it to use community blacklists on their Dovecot/Postfix server. It's not that hard.
Google somehow wants to force me to not use my email server. Fuck them
It you don't have dmarc enabled that is for sure worth doing too.
It does mean your traffic is centralized through a data center, but it probably goes through one already...
Also you're incredibly alienated if you think a "normal person" could manage their own mailserver. 95%+ of western people couldn't use a CLI.
> Spam filtering is a a very difficult thing to do well and Google's implementation was very good
How can these two sentences be true at the same time?
Spam filtering is very hard. Therefore there are some content based rules and some sender based rules. Google is very good at both of those, which means in this context means accuracy and precision based on content and strictness based on sender. Sender based rules make it hard for spammers to send mail pretending to be from a domain, but does make it harder for anyone to send mail. Hosting your own mail server means hosting the authorization architecture to prove you're not a spammer, which makes it harder.
No. It is the domain. No one else in this business has a short pronounceable *mail.com
Except for, oh, I don't know, maybe https://mail.com/ ?
Note: Your browser version is outdated. We recommend using the new Firefox Browser. Download now for free!
the implementation however was much better than hotmail or any other web-based email at the time.
To argue the name was the important aspect is to argue marketing is more important than the quality of what is being marketed, an idea that HN is generally not very open to.
gmail became a brand nearly immediately via initial scarcity of invitations. Hotmail and yahoo mail sounded like idiotic kid email addresses. Webmail never tried doing email. Pobox.com did only forwarding and was too linked to the physical mail in the mind of people. Same went for mailboxes.com.
> To argue the name was the important aspect is to argue marketing is more important than the quality of what is being marketed, an idea that HN is generally not very open to.
Yet the only successful companies are the companies with great marketing and an OK ( or more product ).
yes, to us (the techies), but there are billions of people in the world who didn't know who or what gmail was for a at least 4-5 years. I specifically made the observation of geemail sounding like email because I had that experienced less than a decade ago giving the address to a dentist's secretary in Denmark.
>Yet the only successful companies are the companies with great marketing and an OK ( or more product ).
first of all I'd say Google had a great product and ok marketing for a few years, nowadays they have great marketing, an ok product, and built in market dominance.
So I suppose I can accept that good marketing is a prerequisite but not sufficient.
That said what do you mean by successful? I mean great marketing being required for successful makes you think that the only successful companies can be ones you've heard of, because how successful can a company be that you've never heard of.
And on that note when I tell people I used to work for Thomson Reuters and nobody knows who that is I suppose this means Thomson Reuters is not a successful company?
I would say you have to look at the alternative names - hotmail? yahoo.com? gmail sounds like a fantastic choice.
> That said what do you mean by successful? I mean great marketing being required for successful makes you think that the only successful companies can be ones you've heard of, because how successful can a company be that you've never heard of.
Your target customers know who you are.
> And on that note when I tell people I used to work for Thomson Reuters and nobody knows who that is I suppose this means Thomson Reuters is not a successful company?
Random people are not customers of Thomson Reuters and Thompson Reuters is not in a market of converting them. Those that consume news content as a part of their product know who Thompson Reuters is.
I guess if that is the requirement I would say adequate marketing is required to succeed.
It is not about centralization but about one player winning and beating the competition in that case.
Her actual conclusion is that in order to create a market whose average members are free, we need more regulation.
In the olden days, any business could say in good faith "I want to offer email services" and they would have been treated pretty much the same as any other good faith provider of email service. But today, a business who wanted to start up an email service is insane; just check out the nearby thread of a person who replaced Mailchimp with a semi-self-rolled service. As a matter of fact, no matter what the RFCs say, people in the industry do not treat email as decentralised.
This is partially because people joined the network in bad faith, but it's also because a small number of large nodes are able to assume anyone who isn't known to them is acting in bad faith and therefore force people towards the larger nodes.
As I said before, email and the web continue to be usable in a decentralised fashion, but to do so involves wearing a hairshirt. This is the failure of decentralised applications. The author of the article has a view that is closer to "the market is only as free as its weaker members" rather than the view more fashionable hereabouts "the market is only as free as its stronger members", which leads to a different conclusion. It would probably lead to more insightful disagreements to discuss this underlying difference rather than arguing about what technologies email is less centralised than. (Edit to clairfy: These things in quotation marks are not quotes. The quotation marks are used to help delimit the propositions.)
So the author seems to take "decentralized" here to mean "fully decentralized and peer-to-peer", which is indeed an unconventionally strong definition. But later on, they mention IRC, which is centralized by any definition... Actually, the author doesn't seem to be working from a strong definition at all and they seem to have given this very shallow thought and research.
I must correct you here on "Protocols like Mastodon". The protocol that underlies the Fediverse is the W3C ActivityPub Recommendation [0] and making that distinction is really important. Not doing so would encourage centralization in the same ways as discussed elsewhere in this thread for Email vs. Gmail.
Mastodon is just a fediverse app that was an early adopter of ActivityPub, and managed to become quite popular. Part of this due to a better productization, and the jumpstart this gave them offering the best features first. To many newcomers Fediverse === Mastodon, but - while understandable - this is massively shortselling the interop capabilities of the protocol.
What has already happened by Mastodon dominance is that they are driving the technological direction, but not necessarily in the best possible way and to the overall benefit of the Fediverse. They have a custom client API for instance that is becoming a de-facto standard, adopted by others out of convenience.
More details here: https://overengineer.dev/blog/2019/01/13/activitypub-final-t...
"To clarify what I mean by decentralised: applications whose main purpose is fulfilled as part of a network, where that network is not reliant on any preordained nodes. Decentralised applications are also known as “federated”. There is also a more specific term “distributed”, also known as “peer to peer”. I won’t cover the distinction with the more specific terms because it’s not relevant here. I use the word application in a looser sense, to also cover protocols and specifications."
Even so, IRC has no business in this list per the authors definition.
The "network" part is arguable, but IRC does not require any pre-ordained nodes, which matches the author's definition.
It's a small part of full decentralization, but an important one. Being able to point your Facebook Messenger client at a different server would be a big step.
Interestingly, you used to be able to do the inverse: point your XMPP client at Facebook Messenger. I'm pretty sure they never supported federation, but you could connect clients to multiple servers and have a relatively seamless experience from a single client.
Conversely, in the blockchain space anyone can participate in the blockchain. If a miner refuses to mine your transactions, that's profit that another miner can pick up. And if worst comes to worst, you can set up your own mining infrastructure and mine that transaction yourself - if the system is working correctly, you never need permission from someone else at all to participate. (there are some assumptions here, and if you choose to reject those assumptions as valid you can reasonably claim that blockchains aren't decentralized either)
I'm confused, because it seems like you agree with the author. She says
>> My statement that email can be used without involving Google is technically true, but anyone they blacklist will quickly realise that they’re largely locked out of the global email network.
She is stating that the problem with email isn't that it is an open network, but that it is effectively centralized by a few big players. It might as well be a centralized service. The protocol may be open but if you aren't using a major centralized service you're going to experience a lot more difficulty, including potentially being black listed by these big players. Just getting accidentally classified as spam on one of them and you will locked out. Not to mention dealing with all the spam if you roll your own.
> If you become unhappy with them, it's easy to move.
I think this is an absurd statement and you know it or you are lying to yourself. Sure, forwarding exists but that's not exactly moving. If you've ever tried moving email addresses you'd know how difficult this is. It's significantly harder than moving phone numbers, which also isn't that easy. Just think of every account that you now need to modify. Sure, you don't need to port every account, but there's more than you probably think.
My current provider started getting flaky when they were bought out by a larger one, but I've been putting it off for years trying to move. Maybe offlineimap to a local box, probably best to have your own backup anyway.
Whether or not its possible for one party to dominate the network is one of the potential properties of a "distributed" system that people cannot agree on if its required in the definition.
For example, napster, TOR and IRC are arguably "distributed" systems where there is a central party that controls the network.
IRC is a distributed system much like a CDN is distributed, that is spread across the network.
what we are actually interested in is distributed ownership.
email is that, as is the web, but IRC is not
Nonetheless, that doesn't make it any less of a problem that the definitions are severely overloaded. The issue is not that the article is unclear in its definitions (even though it is imo), the issue is that its hard to have discussions about the topic and compare to other articles, because people talk past each other. Words having agreed meanings is important so we can connect ideas to a broader context and not just talk about individual articles in isolation.
distributed ownership, aka federation does not require trust between operators as each operator is trusted only to manage their own servers without being able to control entities that originate from other servers
Email is (at the server level) a peer to peer network, only tied to DNS/BGP as the only central points, with a global namespace. That can't be said for a lot of open standards.
But, yes, that doesn't mean you have to run your own node. However, it should include the possibility to do so.
i can't see how i would host my own mail server with my domain, but then send emails from that domain via someone else.
the only way i could see that working is to use the ISP/other service as a relay to forward my mails.
I don't like cryptocurrencies (those who are unbacked and resource hungry, I like the tech behind) but bringing payments in the software world and out of financial gatekeepers could exactly the missing piece of the puzzle to have even more working decentralised apps.
Recognising this doesn't help with criticising capitalism, I guess.
To me, "decentralized" doesn't mean every user has to run their own node. It means it's an open network, anyone can join, and there's no centralized control / censorship. You can buy a domain and use Proton, Fastmail, or any other provider to have your own email. If you become unhappy with them, it's easy to move. We have open specs that allow every service to interoperate, so you don't need to know or care what provider your recipient is using.