@bonifartius@qoto.org What did he do this time?
@tyil nothing, sourcehut is getting ddos and it just is a good opportunity to move away from a service which reserves the right to delete accounts for political opinions :)
@bonifartius@qoto.org Ah, fair enough. I don't know if self-hosting Sourcehut is a lot of work, I do like the platform, so I can understand why people choose to use it. Its definitely an easier sell than Gitlab or Github, as Sourcehut is at least completely free software and uses standardized federated systems for communication.
@tyil it's a shame because in principle sourcehut is pretty nice.
self hosting is probably a bit complicated from what i have gathered over the years, it's somewhat of a microservice architecture.
@bonifartius@qoto.org Wonder if there's a market for hosted Sourcehut instances. Clearly it wouldn't even be a big market, as Sourcehut isn't really "dominating" the code forge markets, but just a couple hundred nerds paying 5 bucks a month could sustain a nice side-hustle.
@tyil from what i gathered over the years, it was hard enough to get people paying for sourcehut :) i think hosting things is a very shit business now, as people expect stuff to be working flawless all the time - doesn't matter that the big hosters have more outage than most small things.
hosting things is a very shit business now
It always was tbh, customers complaining all the time because they changed settings they shouldn't have and now some support employee is getting cursed at for hours.@Moon
wanted to reply yesterday, then something happened and i forgot xD
interesting examples indeed! the use case i came across while work usually was "linux vm, only more expensive" :) i was aware of the horizontal scaling, but never have seen an example of "serverless" which made sense.
> This is not suitable for all uses.
i think this is one of the core issues. one has to know most of the pitfalls of the business case and also all the available technologies and their pitfalls. people who are really good at both, or organizations where both are communicating well are pretty rare from what i have seen. might to a degree be a problem of germany though: tech people chase the latest trends but are too conservative to go all in and reap the benefits so one ends with the "expensive vm" solutions.
i was lucky enough to not having to do modern frontend webshit, but what i am told is absolutely horrible.
backend is ok, but i think it's pretty boring that everything is on top of HTTP. i get that HTTP being stateless and short lived connections is pretty good for the mobile world behind NAT and shit, still feels bad that things are this way. it's great to at least use a language one likes. i still have to give elixir a try :)
I was able to build small systems that only activated when there were requests to the system by utilizing cloud services, which only accrued storage costs on a monthly basis. And then a small amount of processing cost when actually doing work for someone. As an example for my bosses to demonstrate the power of this, I built an online forum system that used "serverless" functions that were called when you POST to a website URL, which inserted the message into a NOSQL database, then rendered the page to an html document into an S3 bucket, which was served behind a free-tier CloudFlare web URL. You didn't have to run/update/patch even a single server for this and you only paid S3 storage costs and the cost of running the serverless function, of which the first several thousand requests were completely free. This resulted in a full web forum that cost pennies a month to run.
On the large-system end I was able to build clusters of computers that would grow and shrink on demand. The per-CPU cost was higher, but it saved money overall because it automatically grew the number of CPUs under load and then shrank again when the load went down so you didn't have to just provision the maximum number at all times which is wasteful, and if your prediction is wrong and you need extra capacity in a moment's notice after all, you have to go to cloud anyway.
This is not suitable for all uses. If you have a very good idea of how much and how predictable load your system is going to have then you can pay for a fixed number of CPUs or run your own CPUs on-premises and it will cost you a lot less than paying for metered CPU.
With regard to webshit, I have done that too and in a large company yes it's a nightmare. If you're in a small company and setting your own practices and deciding on your own tools it still can be awful but it's a lot better. Example, being able to pick a non-shitty alternative to webpack for bundling, being able to standardize on TypeScript, etc. Things are always better when you can choose your own tools rather than have them set by a company committee to accommodate your lowest-skilled employees.
However, there is also, simply, non-frontend work! Today I only do backend. I am standardized now on using Elixir programming language for development and my job is lovely.