"Throw away" base OS and implications?

[mapleLC]

I'm looking to setup a retirement work environment which includes an EE lab, software engineering lab, home office, and personal.

One of my software engineering explorations is setting up a distributed computer system that runs and hosts, well my life, basically.

A major problem, even graduate it to a social problem, is data whore who cares I'm not hiding anything data whore mentality.  They lulled the proles into believing since they weren't hiding anything, there was no reason not to let them invade privacy to astonishing levels.

So, being averse to this kind of whoring of my alter-ego, I decided to distribute everything internally, at home.  No more skyping eachother at home at our desks, we have hosted IM now.  This kind of thing.

This is a lot of preface, sorry.  Anyway, with this type of "hosted internally distribution" in mind, all of the utility of my desktop computer is gone.  I just need a clipboard to bounce around the different applications I use.  So a browser will do a lot of the lifting, but then I need a way to remote into virtualized environments.  There is nothing I can think of my desktop needing to do other than browse and remote desktop, then have settings to maintain itself.

So here is the question, coupled with a thought experiment:

Has anyone every heard of an OS like I'm describing and if not, what is the closest approximation you can think of?

In terms of a thought experiment, help me take this topic of internal distribution of applications to its logical end, which helps flush out architectural considerations for me early and may perhaps trigger others to consider similar ideas. Next post will have a basic outline of network infrastructure to push that along.

[mapleLC]

Network infrastructure:

WAN > DMZ > Router 1 and 2 > Router 1 LAN < Router 2 VPN-LAN

SAN storage

Proxmox cluster for hosting OS and apps:

• 3 Windows desktops virtualized and repurposed as proxmox node 1, 2 and 3
• Hosted VMs for various OS - base Ubuntu desktop, Mac OS ventura, etc
• Docker
• BookStack

BookStack was an important first one to get in there, so there I can accumulate all the notes and install info, otherwise its back to browser tabs or bookmarks.  Its fundamental to moving away from a desktop since BookStack replaces a lot of the simple things a desktop would do.  Think of it like this: Windows > my documents ... thats on the SAN now... I can run a document editor from a browser, I dont need local storage, and I can organize the file in my BookStack notebook which I can log into from anywhere (in the world, if I want to port out).

I started by setting up an Ubuntu desktop, installed bookstack.  I started the virtualization process, notated it, then converted 3 desktops into my Proxmox cluster, the boxes are stood up but the cluster is not complete.

That BookStack is now virtualized.  Its kind of a notebook, for lack of a better explanation.

Here is kind of the green field... I can basically setup anything at this point.  What have you done, would like to, or suggest trying?

All the areas are open to tweaks, DMZ, routers, setup, etc.

[mapleLC]

I'll keep my research here as a record in the event nobody takes an interest in the topic now, perhaps in the future it might have more meaning.

https://tails.boum.org/index.en.html

This is quite an interesting operating system.  It does not tick many boxes, but for security research its really useful.  It has a lot of simple tools, not the sharpest look and feel but its purpose built.  I am going to test it as a "browser" base, so I am going to isolate my browsing there for a period and see how functional all the normal things I use are. 

[shapirus]

What you're describing is known as "thin client". It doesn't even need to have a disk drive (and use network boot instead), so no local storage.

[alm]

Exactly. A thin client could either boot from network or from a CF card. The boot medium would be read only (except occasional new image with security updates) and the only persistent storage would be on network servers.

Could be Linux-based, for example https://ltsp.org/docs/installation/#maintaining-a-client-image or https://thinstation.github.io/thinstation/. You can also make them with Windows, but I think you'll either have to create an image yourself from scratch or buy an off the shelf thin client with hardware and software.

[mapleLC]

What you're describing is known as "thin client". It doesn't even need to have a disk drive (and use network boot instead), so no local storage.

Exactly. A thin client could either boot from network or from a CF card. The boot medium would be read only (except occasional new image with security updates) and the only persistent storage would be on network servers.

Could be Linux-based, for example https://ltsp.org/docs/installation/#maintaining-a-client-image or https://thinstation.github.io/thinstation/. You can also make them with Windows, but I think you'll either have to create an image yourself from scratch or buy an off the shelf thin client with hardware and software.

Actually, it's an apt description since thin clients are a hub and spoke model and what I am exploring shares similarity.  The main problem is graphics, TCs are not up to the job.

[alm]

So build your own thin client. No one says a thin client can't have beefy graphics card. Of course making use of the graphics card over the network could be a bit of a challenge.

[mapleLC]

So build your own thin client. No one says a thin client can't have beefy graphics card. Of course making use of the graphics card over the network could be a bit of a challenge.

I'd need a machine near me, a SFF is a little larger and generally has a PCI slot.

Look at this, smallest OS I have seen and runs x86.  40 megs!

http://kolibrios.org/en/

[IanB]

The classical thin client implementation many years ago was X. Your "thin client" machine would host an X server, and potentially nothing else, and the remote machine(s) would display their UI inside an X window on the user's local desktop. We used to do this back in the 90's where the all the compute power was on remote centralized machines and the desktop client machines had lower compute power. The idea was that each time you launched an app, you would tell it on which remote machine it should display its windows.

I have no idea whether this technology is still relevant today. It is similar to remote desktop apps, but much more flexible. It is an example of where computing technology often regresses over time rather than advancing.

[ve7xen]

I have no idea whether this technology is still relevant today. It is similar to remote desktop apps, but much more flexible. It is an example of where computing technology often regresses over time rather than advancing.

You can absolutely still do this today, usually via SSH forwarding, and it's still very useful. Most 'heavy' applications these days are not efficient in their use of X11 calls though, so it can be quite slow, but it's a lot easier / more secure than most remote desktop options to get working. For example

Code: [Select]

ssh -Y desktop wireshark
As far as the OP's ask, the closest thing I know of out-of-the-box is probably ChromiumOS. Otherwise I'd either go with a lightweight 'normal' distribution like maybe Ubuntu MATE, or if you're more the 'build it myself' type, maybe NixOS is a good fit here.

[mapleLC]

As far as the OP's ask, the closest thing I know of out-of-the-box is probably ChromiumOS. Otherwise I'd either go with a lightweight 'normal' distribution like maybe Ubuntu MATE, or if you're more the 'build it myself' type, maybe NixOS is a good fit here.

Not a chance I'd touch anything Google.  That company has become a social problem.

I have not heard of Ubuntu MATE, will look.

Man, I wish I wasn't alone on an island with privacy, these prices are ouchy...

https://www.raptorcs.com/TALOSII/

[shapirus]

I have not heard of Ubuntu MATE, will look.

fyi, it's the same ubuntu as any other. the only difference between all those flavors of ubuntu is the desktop environment that is installed by default. you can turn one into another without reinstalling them, by simply installing the respective packages using apt.

[rhb]

My NCD Xterm was so nice that when I came back from vacation and found my NCD had been replaced  with a SPARCstation1+ I demanded my Xterm back.  My primary job at the time was porting stuff during the 'workstation wars' of the mid-90's.  The 1024 x 1024 grey scale screen let me reproduce at full size a letter size image of the code.

These days an obsolete PC with a good graphics card should serve very well in that role running a heavily stripped down copy of Linux.

Good luck,
Reg

[DiTBho]

look at xrdp
I am studying it to solve a similar problem.

[DiTBho]

My NCD Xterm

NCD Xterm << Tektronix Xterm 

[mapleLC]

look at xrdp
I am studying it to solve a similar problem.

I use it now to remote between win and linux machines, what do you mean similar problem?

[DiTBho]

look at xrdp
I am studying it to solve a similar problem.

I use it now to remote between win and linux machines, what do you mean similar problem?

on x11 when you export the DISPLAY you have a whole series of overheads due to the x11 protocol.

I am developing an xterm-machine (based on ppc40x), xrdp accepts connections from a variety of RDP clients, and RDP transport without encryption makes things very fast when you connect to a GNU/Linux desktop using RDP from xorgxrdp Xorg.

xorgxrdp + xrdp + x11