My point of vue is CasaOS / Unraid / Umbrel / ... serve a good "first base" with selfhosting. Kind of like a gateway drug: gives you the candy to see how nice it could be but really under the hood, they are lacking a lot of substance.

I would never advise someone to limit their experience to those tools thought, as they lack so many things that are required for a proper long term selfhosting setup (monitoring, backups, encryptions, reverse-proxy, etc..............). It's a decent start thought.

Finally one criticism I could make is, unlike what you often read, I think it's ok to abstract things. But the issue is, if you're going to abstract away Docker completely you better make sure to offer everything the user needs to deal with their apps, and as far as I can tell, not only it's not the case, but also those tools kind of tend to be opiniated in questionable ways. I have never used CasaOS thought, so it's only 3rd party observation

You need to setup the hostname in the Cosmos installer if that's what you are asking. You can put your IP or something if you dont have your domaoin yet

- I think domain is preferable for home servers because then you get subdomains for apps, which are easier and can also share the auth cookies for SSO

- you probably had a cached certificate

You dont need to do anything to migrate, Cosmos will just work with Portainer, including just picking up your existing containers

You are correct, and Sab is due to be added

Docker is an important ingredient in the mix, to isolate the applications completely, and make things much more streamlined than traditional VM, but I understand if it's not for everyone

Truenas

Haven't used it, but it looks like there are overlap.

Cosmos does not yet have storage management (but soon) and uses Docker instead of VM

I am considering Podman support but probably more next year when Cosmos is feature-complete for 1.0

Keep in mind it might be a challenge to do everything rootless but I will see what I can do

You need root access to manage docker containers that's (almost) unavoidable. Also Cosmos does not support managing remote docker instances. On the other hand, a good (and secure) pattern is to use Constellation (the integrated VPN) on 2 servers with cosmos installed on each. you can connect them together. One of the servers (the seedbox) is the main server running services but it is not exposed on the internet and the only way to access it is to connect to the VPN on the other VPS

Docker is an important ingredient in the mix, to isolate the applications completely, and make things much more streamlined than traditional VM, but I understand if it's not for everyone!

Don't get me wrong, I am fully aware that you need to reduce as much as possible the amount of access something has but as you said:

you should never have permissions to things you don't need

well Cosmos needs to see your files if you want Cosmos to manage your files. It's that simple. By default its on because it is needed for Cosmos to function. You can remove it, but at the expense of some of the functionalities of the server.

By the way Cosmos, as a Docker management software, has access to your docker socket. Which mean, you can remove anything you want from the container, technically, it can add it back itself. Having access to the socket means being able to manage the containers, including itself. In other words, having this mount in the docker run command is just a comfort thing, but in term of privilege, whether it's Cosmos or Portainer or any other docker manager, they have full root access to your system and that's unavoidable.

why not have -v /CasaFolder:/mnt/host or something similar

Because it would require users to always update their Cosmos containers to add additional folders all the time, giving a terrible and very error prone user experience.

If there is a solution out there, that solves that problem (as in allows Cosmos to continue to work the same without that mount) then I will gladly implement it. But as far as I can see there isn't such solution

I think you've stumble accross few of the huge issues with selfhosting

- Developing apps is too hard, you have all the difficulties of SaaS development but with the added difficulty of having to support people installing your app in various setups

- For the difficulty, the return on investment is low because the community is much smaller than what you can touch with a SaaS software

This causes the breadth of available apps to be quite shallow, and additionally, another factor threaten further that diversity is that

- people gets into self-hosting in one of two ways. Either to create illegal media-center (in which case they install Plex, Jellyfin, *arr, download client, etc...) or to manage their document in privacy (Nextcloud, etc..) seems like you are type 2. This causes most projects to focus around those hot topics, without exploring other things (this year alone at least 4 photos albums backup software started development..)

But this state of affair is not sad or inflicting, it is natural for such as a young community to take time to find itself, especially in this difficult setting (I know selfhosting is not new, but I call it young because only recently did it start becoming so popular). And there are solutions to those problem too. On my end, like many other talented people, I am working on technologies to improve this situation, and hopefully one day we will see a large diversity of application growing, with much more accessible setup for people to run.

What I forsee will be big in the future

- Once we crack federation (I do not think current state of the technology is good enough) social app (Video sharing, file sharing, social media alternatives, news site etc..) will be big

- Going back to news, once we improve the QOL of SH for public sites, news agglomeration is going to be big as well (for blogs and stuff)

- Any mobile/SaaS app could have a SH counter part, that will automatically gain benefits from not being in the cloud. Im thinking things like various task management, productivity tools, and of course, home automation is gonna be the bigger winner for being in the home already, therefore workable offline. An example of this is already happening with cooking/recpies apps (Mealie, Tandorii, Grocy, etc...) which benefit from being at home, private, and accessible from the family, and home-assistant.

- Finally, SH is going to supercharge the development of very niche software. It makes no sense to develop an entire SaaS offering for 100 users (ex. a software to manage your model train would be very niche) because you have to pay for a domain, servers, and so on... But a SH app could literally cost $0 to run (for the devs) while yelding minimal benefits (either from subs or donation).

Give it 2-3 years for those stuff to develop better. In 3 years this sub will be almost twice as big at 500k, and you will have 2-3 times the amount of apps available that's pretty much a garantee