Hacker News Reader

" I think at that scale you start to care about things like uptime and maintenance."

I guess none of members of your mentioned groups will ever grow to care about a scale. And thats a good thing because they are distributed. And i think it creates a better internet, which is kind of wide network instead of shallow graph of couple mega nodes.

The biggest win we could achieve for self-hosting doesn't involve people actually self-hosting. The idea of personally maintaining your own infrastructure is unlikely to scale to general population - but the thing we're really after is ownership of data and the ability to own infrastructure.

So, I think, the ideal situation would be to have a combination of big and small companies offering storage and compute as a commodity. You'd pay a fee to keep your stuff hosted somewhere; any time you see a better offer, you can migrate to a different provider without much hassle, and with near-zero downtime. Cloud services would work by shipping their code to your data, not the other way around[0]. And if you were so inclined, you could just build your own infra, or even buy a turn-key "self-hosting in a box" kit.

Pieces of that vision are already here. Compute providers are plenty. You can order "self-hosting in a box" kits. Internet architecture makes everyone's computer equal (at least in theory, ISPs mess it up with NAT, and their T&Cs). The only thing missing is the part where you own your data, and SaaS vendors serve you - the bit that makes SaaS truly be Software as a Service, instead of Serfdom as a Service.

--

[0] - Preferably with homomorphic encryption preventing SaaS vendors from putting their hands in the cookie jar, if we can get that to work without creating another blockchain-level environmental disaster.

You mean good old webhosting? What I don't get about the HN crowd is making every simple, already existing, already proven, and already solved problem so damn hard. I'm in EU, I host my websites at multiple local webhosting companies. They're small enough to care about support and big enough to guarantee speedy and reliable service. By law they're not allowed to go through my data (ofcourse they can and I have no way of proving that they did), so the legal deal between me and them is crystal clear. Who do you call when your Amazon Web Shit serverless thingy doesn't work no more? I can and did move several webapps and websites from one webhosting company to the other. It works flawless. And besides waiting a couple of hours for a DNS change, it's almost instant. I get it, you can't do that easily with a system with million users. This so called problem was already solved decades ago. It was called personal computers and the internet. When people started calling the internet 'the cloud' then things went downhill. Excuse me for the slight rant. I should go outside and see some more sun. ;)

Yes, this is solved. I'm saying that there's another problem that needs to be solved too: control over data.

Regardless of what you use for hosting your own stuff, if you want to use a third-party SaaS as a user, they own the data. Want to make a document on Google Docs? That document lives on Google's servers, it's forever tied to their service and mined by them. There's no artifact you can hold on to, other than your user account.

What we need is a system where the data for that Google Docs document lives in a place you control - be it your own hardware, or some hosting you rent somewhere. It's the SaaS that should come to the data, and operate on it there. That way, if you lose your Google Docs account, or decide to edit the document with something else, you actually have that document, in its canonical form. Same for all other SaaS.

Universal formats?

In an ideal world, yes. In practice, that's impossible to do up front, so the next best thing would be open formats - i.e. openly and fully documented ones. The goal is to break the leverage a vendor has over the users when using a closed, proprietary format.

Where do I start to help?

> There's a whole economy in making it difficult for these!

And there is your niche right there. Individual companies or persons do not need to make 'dents in markets' all by themself. A host of people doing the same might. But for the individual - especially when having sustainable income objectives, not hockeystick growth - there's a good place in the market I think.

> Hypothetically in a future where self hosting is non hostile, will you see people self-hosting startups and the like?

Huh, is that not a thing anymore for startups today? In my experience, starting in 2003, across two companies, was that if you need any service fast and on the cheap, you're forced to host it yourself. Run your own mail server, web server, DNS, server housing, all the web apps, etc. and of course spend most of the day developing your actual product. Have the hosted options actually gotten so cheap and reliable these days? Is it a mindset thing?

My gripe with the cloud or SaaS solutions I've used over the years is always that management of them and backup is difficult tending towards impossible and without those you can't rely on these services, it's just voluntary vendor lock-in. When self-hosting, your actually (forced to) learn how they work and be able to fix them. Self-hosting, to me, is the simpler, more reliable option, if you depend on a service for your business and it enables you to get it fixed yourself, when it is not working. It doesn't prevent you to out source the work, if you have the money, but worst case, you still have direct access to your property.

With the hosted options, in my experience, you get all these fancy promises on availability and it being the latest, smartest tech and then the service is down for half a day, your data got restored from days old backups (if at all) and there is nothing you can do except tell your customers that "were sorry and working on it" while you wait for it to come back. :-(

>and then the service is down for half a day, your data got restored from days old backups (if at all)

It's not unheard of for the same thing to happen with in-house systems.

The difference is that you may have a wider range of options to avoid and respond to any outage if you run things yourself. That seems to be a rather theoretical advantage though. Every time there's a new ransomware attack, it's always those in-house deployments that are hit hardest and take the longest to recover.

I think under optimal conditions self-hosting is superior. But conditions are rarely optimal. As soon as you have to convince non-technical management to invest in non-productive necessities or in contingency planning you're already in a sub-optimal position.

It's economically valid reasoning but most startups (even one man band) are happy to pay a premium for the cloud.