Mail, and running your own mail, is not hard at all when you're doing it on an individual personal mail scale and not sending spam (or any other kind of bulk/automated mailings you want to pretend aren't spam).
The whole system is a few hundred lines of C with no dynamic object lifetimes or variable buffer sizes.
Oct 31, 2026, 8:51 PM@dalias I'd like to see it! how you receive mail, and how you send it. This is probably very different from what I consider an MTA should do, and your data point would be valuable.
@ska How I send is with mxclient, but I have a wrapper that backgrounds and retries according to schedule. No crossing of privilege domains; it runs as user. I do want to enhance mxclient to support SOCKS and ssh -W transport.
How I receive is with a tiny program that parses SMTP (including BearSSL for STARTTLS, which is mandatory per DANE) to determine the user to deliver to and pipes the mail contents to the user's configured acceptance program, which can dynamically produce temp or permanent failure (which the sender sees at SMTP time) via exit codes.
There is 100% backpressure everywhere in the system, no queues, no backup MX's, no MITMs. During down time it's the sender's responsibility to keep the mail and retry later.
@alwayscurious @ska Yeah, I'm not particularly interested in the NSA's latest backdoored replacements for crypto so 🤷
@alwayscurious @ska "For something to be a backdoor, you need some way to ensure that nobody else can exploit it, otherwise it is not a backdoor, but a broken algorithm"
This is semantic pedantry. The NSA DGAF about "NOBUS". They are happy just making everyone insecure as long as they can be one of the parties intercepting.
Any attempt to induce churn into crypto algorithms and implementations and displace ones that have stood the test of time protecting trillions of dollars of supposed value with things that have only been used seriously for a short while or not at all is *sabotaging* our security. Whether you want to call it a "backdoor" isn't really important. It's bad.
@alwayscurious @ska In the case of the current push for PQ, there is another nefarious angle too: pumping the valuation of quantum computing scams, which are the next thing the former coinbros, now AI bros, are planning to pivot to once their AI thing blows up.
@alwayscurious @ska He is one of the AI cultists who need to pivot to the QC scam ASAP.
Nov 1, 2026, 10:34 PM@alwayscurious @dalias @ska He's also been part of the cult which yielded the current AI labs since the 2000s. His lesswrong account was registered ~2mo after the site was founded.
He's also a weird Zionist and mostly notable (outside of academia) for stating that women never liked him much, therefore misogyny in his field cannot possibly be a real problem (https://scottaaronson.blog/?p=2091#comment-326664). But that's neither here nor there I suppose.
Nov 1, 2026, 10:55 PM@ne_ @alwayscurious @dalias @ska He's also got a bunch of other pretty awful opinions besides. I pretty much noped out from him once I realized his unapologetic Zionism, but his blog is a goddamn _constellation_ of bad takes on almost every topic.