So I was researching different distros, and I stumbled upon one called Poseidon (based on Ubuntu) which was intended primarily for scientific modeling. However, it hasn’t been active since 2018; also, after Poseidon 4, they shifted to focus primarily on oceanography.
So my question is, firstly, is there an active distro that has a similar intended purpose, and comes with all the relevant software? Barring that, is it easy enough to replicate just by downloading the relevant packages?
If the answer to both is no, my question is, how could one best go about resurrecting it as a fork? Would it be better to start from either Poseidon 9 (the latest) or Poseidon 4 (before they shifted direction), and then try to update all the core components? Or to start with the latest Ubuntu (or better yet Debian), and then simply install all the software needed to make it functionally identical to Poseidon?
Bonus question: if you start with an OS, and gradually replace one component at a time until all components have been replaced, is it still the same OS? (Theseus was a son of Poseidon, but unfortunately the name is already taken)
There’s not much info about it that I can find. Do you have more info about what was specific about it? Wikipedia says:
Poseidon Linux was aimed at the international scientific community, it came with a range of applications for fields such as GIS, 3D visualization, and statistics.
So it sounds like it just came with some apps pre-packaged (which is what most Linux distros do). If that’s the case then yes, you absolutely can. How easy it is depends on what the specific apps are and their installation methods.
Yeah, it seems like it can be easily replaced insofar as comparable software exists for other distros
If you are ready to build your own packages and host your own repos, go for it. You should have some form of automation to at least get security updates, and it will take some recurring maintenance time no matter how you go about it. I haven’t looked closer at Poseidons repos but I think if this is a good idea or not depends a lot on the state of their build/packaging/distro code. It could be just a matter of cloning the repos, changing a few config parameters and running a couple of commands, or it could mean significant work, depending on how well it is engineered. You could start out and if you find yourself digging to deep in the process of getting a build, back out and reassess.
Kind of like others suggested, there are lighter options. All the big dists should have some form of build tooling to make kickstart/preseed/spins/whatever they call it where you can prepare a custom install ISO with your own set of packages. There are tools like packer and Nix that you can use either as part of building OS images or just script to run on a clean base installation of some dist. You can make an ansible playbook to automate setup and have that run by cloud-init. You could make a shell script to automate the installation of packages and setting up of environment.
maybe a distro like this is less needed now that we have Flatpak, AppImages, and (yuck) Snap
you mentioned 2018, I think that’s about when Discover added support for Flatpak and Flathub
That makes sense, it seems like the best bet is just to find software for all the same functions
I’d rather use a container for reproducibility
When doing science, reproducibility is a test that (in)validates hypothesis. If your software has a bug that creates a pattern in your analysis, we should hope that this bug does not happen in your collegues pc when they try to reproduce the analysis.
Good thought.
With containers, you can tedt the code with your container and with mine. If you only try yours, you have to wait for me to compare the results. With containers, I shio my build and you can check the results immediately.
Yeah, thats useful too.
All Linux distros are roughly the same under the hood. They have a version of the Linux kernel, a service management system, a basic set of packages and preinstalled software, and some kind of UI. How those pieces are put together realistically doesn’t change much. What distro maintainers then do is update the packages and kernel and make sure they work together.
Specific design distros like the one for scientific modelling is probably based around having a lot of data processing stuff built in. Unless they had a specially modified kernel (which is rare, but possible if they had to integrate special lab data collection hardware support), I see no reason why you couldn’t just recreate the same package structure on a modern Linux 6 kernel and have it be a spiritual successor to the original distro.
Sweet! That’s what I was hoping to hear. Seems like the simplest solution. Thanks for the info.
Should I mirror an archived version onto a VM to view the package structure and copy it on a modern distro? Or is there a simpler way to see what packages it would need?
I’m new to linux, so forgive me if it seems like an obvious question
With the caveat that I’m just a random idiot on the internet who has never even considered building a custom distro — seems like the most reliable way to go about it would be to build such a thing on top of a declarative distro such as Guix or NixOS. It should require less maintenance effort to make sure all packages are compatible as they’re updated, in case you decide to publish your work. Nixbook is one example of this. Someone on Reddit was also making a case for reproducible builds in scientific computing. A Docker container could also work if you can do without a GUI.
Seems like the Poseidon Website has a PDF that lists some of the software that came with Poseidon 3.
Since it was based on Ubuntu, I’d probably start by using a modern version of either Ubuntu or Mint and then seeing if any of that software you’re interested in is available for those distros in their software repositories. No clue if any of the more specialized stuff it is still maintained, or if there are better alternatives nowadays. You can search the default Ubuntu repos online without having to first install the OS here: Ubuntu Packages, if you want
Awesome, that list is very helpful. Thanks!
Here’s a full package list for 9; you can get lists for earlier versions, but maybe not so far back as 4: https://distrowatch.com/table-mobile.php?distribution=poseidon
Unfortunately, it’s not clear which packages are included for specialization; if you were familiar wiþ Linux, you could visually filter out þe common ones.
I would suggest you pick a distro based on oþer considerations. E.g., Mint (also Debian based) is often recommended for new Linux users. Þen do some search engine querying for physics software and just install whatever looks interesting.
If you’re focusing on physics, I’ll bet a dollar you’re also going to need a bunch of LaTeX stuff (unless higher ed hard sciences have been surprisingly innovative and started accepting Typst). Þere should be meta-packages which will pull in þe GB of various LaTeX related stuff.
A “distro” is basically just:
- a precompiled kernel
- some preinstalled software
Idk anything about Poseidon but,
is it easy enough to replicate just by downloading the relevant packages?
The answer to that is yes. Just pick a base you like (eg debian, fedora, ubuntu, …) and install the software you need. You could automate it with some simple scripts, or be fancy and write a Butane config to preconfigure a base Fedora CoreOS image. IIRC, the Omarchy distro is just the former, not even a proper “distro”.
If you want to create your own “proper” distro that other people will want to use, there’s a lot more that goes into it: updates, builds, tests, deployment, patch sets (because you’ll inevitably need to patch various components for compatibility), bug reports, some kind of governance structure…
It’s a whole software development thing. If you just need a customized platform for your buddies/workplace, customizing an existing distro is the only reasonable choice. I’d suggest looking into bootc and ublue if you need more than a simple post-install shell script.
Poseidon also had daughters.
To me resurrecting and maintaining a distro would be a bit harder and more work than to just installing the software you need. Could be a rewarding learning experience though.
You could always try to contact the original maintainers and see if they have any advice. Maybe why the project was abandoned.
Yeah, I only joked about Theseus due to the Ship of Theseus analogy (replacing all the parts of an OS)
Overall, it seems like the best option is just to install the software one needs
If you want to do it, use ublue. You don’t have to learn how to create a distro. You just use a ublue base and add the packages you want. It’s incredibly easy
What made this distro distinctive, and for what features?
For academic and scientific use. Based on Ubuntu, but enhanced by GIS/maps, numerical modelling, 2D/3D/4D visualization, statistics, tools for creating simple and complex graphics, programming languages.
https://en.wikipedia.org/wiki/List_of_Linux_distributions
Seems useful, but if it can be replicated simply with analogous software then that’s likely the best option
