I discussed the problems with centralizing the internet (on us-east-1 in particular!) with Lex Fridman just this summer: How we've strayed from DARPA's original design, forgotten how enjoyable and profitable owning servers can be, yet still can be thankful for the cloud gifts.

@lexfridman You mentioned (I think) playing with distributed apps, mini servers, and employee fibre broadband. Did I miss hear..or any further thoughts on this?

@lexfridman I'm working on something very exciting in this domain. Can't wait to share!

Can’t wait to look at this!! A bunch of distributed, open source, essential business apps would be AWESOME (website, email, drive, chat, CRM)! Maybe the source/builds could also be distributed …and rather than AI/developers building the same widget a miliion times, ‘pay per play’ integrated from the get go eg can natively offer compute/BTC for specific features to be built 👍..a boy can dream 😂

@lexfridman We have an app hosted exclusively on us-east-1. Luckily, nobody uses it so we were not affected by the outage. 🤨🤣

@lexfridman People aren't running servers because the committees that maintain the protocols forgot to keep them simple and accessible for people to administer. My (very large) ISP doesn't even want to run its own email server anymore.

@lexfridman none of this is solved by going to a different provider or a datacenter. those that failed when us-east-1 went down would've failed with any other single point of failure solution.

I've been running my own server, in my house, for over 20 years. Sure it's not going to be scalable and I have to pick and choose what to host so that I don't get hacked. But what I'm running doesn't need to be so complex that I need someone else to host it. My next project will be a git remote. Do I really need to play Atlassian$35 a month? Probably not.

@lexfridman Made the greatest thumbnail

@lexfridman In spite of, and based on the original DARPA design, DNS has always been the Achilles' heel of the Internet. I was involved with the IETF group for DNS2 in the early 90s, which never really achieved the fixes that we were hoping to see.

There was an alternate path we could have gone down that makes running your own infra easier and less complex, but we chose outsourcing. That path is still open to us and there are tools that make running your own infrastructure easier than ever, but they need people with strong opinions to advocate and guide people.

@lexfridman Great interview! - worth watching the full version👏 Regarding going back to distributed and decentralized computing, network and storage i would like to point at the @threefold_io project. I am running a 3Node myself in Denmark and renting VM's in EU countries as backup🙌

@lexfridman In 1993, i already complained about HTTP and the need of a server.

@lexfridman I love running my own servers

@lexfridman It's unbelievable to think almost an entire industry of otherwise smart, curious nerds were convinced to give up their hardware. Attempting to solve scalability and reliability by pulling on sliders in a control panel is a mistake. An expensive mistake.

That was such a great conversation. I listened on a road trip from LA to SF and was completely engrossed and felt like I was in the room with you guys. The way you covered family, career, growing/scaling, racing and design so thoughtfully still resonates almost 3 months later. Best podcast I listened to all year, by far.

@lexfridman I mean even though I'm a proponent of this line of thinking, I can also see the reason we are where we are. To some extent it was the price we paid for faster time-to-ship and wider accessibility. But like so many things this decade, this really has gone too far.

@lexfridman You haven't replied to my question about bandwidth and bandwidth costs. how much bandwidth does each of your data centers have and what happens if you need more / at what cost?

@lexfridman Bring back on-prem!

@lexfridman My very tiny company is doing our own on-prem and I am stoked. Yes it will be tiny and insignificant in any serious comparison to anyone, but it is ours and we get to build stuff. Feels like a privilege.

@lexfridman My thoughts were with you yesterday.

Deploying to the cloud is lazy sysadmin. No lift, pay extra. Though it is great for Disaster Recovery for a SMB. A fire or weather event takes out your on-prem hosted DC. Deploy straight to cloud and get sales pumping again. Just don’t be lazy and leave it there. Make sure your plan includes multi region during, and return to on-prem when recovery of physical infrastructure is complete. How to convey this to a non-techie leadership who thinks “it always works right!”is the hardest part of sysdamin.

@lexfridman Is automated bare metal cool again? 🤘

@lexfridman Make Belts AND Braces Great Again!

@lexfridman seams we are getting old... I fully agree, we run all our infra for years on our own servers using exactly the same cloud setup as we would rent from somewhere.

@lexfridman I have a 5 hour cross country flight today, gonna have to watch this and get fired up

@lexfridman I felt that one in my soul 🤕

@lexfridman Being a systems engineer, managing compute in the data-center was the favorite part of my 22'ish year career. Wish more organizations went back to this, it would be wonderful.

@lexfridman . @dhh warned you!

@lexfridman if you run or provision GPUs in your daily, get on @akashnet and decentralize. whether you’re a neo-cloud or an AI startup you’ll save

@lexfridman us-east-1 is actually a threat to national security.

@lexfridman I love this back-to-basics, hacker/tinkerer enthusiasm, it’s infectious!

1. Agree with owning your infrastructure and servers vs cloud nowadays 2. Cloud definitely helped us progress to the point where we can now run our own cloud like on-prem environments. I'm curious to know more about DARPA’s original design. I've never heard of it. I'll be the first to say it's been really great to have a few roles where I've been able to work in the data centers and labs too interface with the physical servers. There's nothing like server fans blowing air at you.

@lexfridman Man was foreshadowing 😂

@lexfridman Based

@lexfridman Exactly why I have an R730 at home...

@lexfridman I still have yet to see the insides of a hardware server :) Always managed them remotely; there's even Dell's DRAC thing to avoid calling the DC when you screw up!

@lexfridman If you don't have a rack of server gear in your home lab, and you don't expose north of a dozen services via reverse proxy and DDNS services to report your IP to your domain provider, do you really even computer?