The only graphical installed I've seen was the corporate 'build' of redhat, a 65 page mammouth document that talked about 'disabling telnet'. WTF? Why the hell is telnetd even on the machine? They also had X on there -- on a frickin server.

That's when I spent some time doing an ubuntu build, which even as of 16.04 is ncurses based installer. Boot from pxe, give an IP address/gateway/dns/hostname etc, approve the automatic partitioning selection, select which support area looks after the machine (for user generation, monitoring, etc), and that's it.

Everything gets the same base install, including things like active directory based logons (with ssh keys for those who want them), automatically adds itself to nagios. Build a dev installation (which you select form pxelinux) and auto patching gets enabled, and syslogs get fired off the the right location.

On a VM this can of course be passed as part of the creation process, so it just take 5 minutes to build from scratch. When it comes to building on real hardware it's just a matter of typing the stuff into the boxes then going and making a cup of tea.

After the base install is on, you can then either apply an ansible script to do various package installs, or do it old-skool with apt-get install $package.

It's entirely possible that debian testing may work, however from a corporate signoff (and when you've got nearly 2,000 of these machines installed you need that) saying you're basing the company on 'Ubuntu LTS' is far better than basing it on "Debian Testing".

We've run this way for over a decade over 6 continents, it just works.

>Are you re-installing, or re-starting, a VM for this?

Yes, of course I am, as an amateur in linux it would be reckless for me reuse an existing linux installation in a different context. The only way I can safely re-use is docker.

>An installation from scratch, in any case, should only ask you hostname, domain name

Domain name has already lost me, I'm not a website, why do I need a domain? :)

In practice, you're usually asked about root, usernames, home folders, etc, too.

>none of this stuff benefits from a flashy / GUI installer

No it doesn't need a flashy installer, but I'm equally talking about the Ubuntu Server edition's text installer. I'm not sure if you're talking about GUI vs CLI or "graphical" vs "text-based oldschool GUI". I would say totally vanilla CLI installers absolutely don't explain things well.

> Yes, of course I am, as an amateur in linux it would be reckless for me reuse an existing linux installation in a different context. The only way I can safely re-use is docker.

If you're using Docker - or even know about it - you're not an amateur.

Since you're talking about your need to 'build a Linux VM for something' this also suggests you're significantly better than average on the GNU/Linux awareness scale.

But to your question -- if you're using a VM instance, make a base build, make it just right, and then snapshot it. When you need to spin up a new Linux VM for something, replicate the snapshot, do an apt-get update/upgrade (or equiv) and carry on. This is fairly common practice and saves having to remember your domain name, root password, home folders etc.

To your last point -- I rarely install Ubuntu, and can really only compare Debian vs various Microsoft installations processes for the past twenty years. GUI GNU/Linux installers were popularised by Ubuntu, and that's why that came up. You're right that totally vanilla CLI installers don't explain things well, but that's what we used to have on Windows 3.1 / 95 / 98 ... and people somehow managed to struggle on. Either average Microsoft users have become dumber (quite possible) or IT types continue to underestimate people's ability to read or work things out. Given almost everyone that installs an OS now has another device to hand to look up guides and answers, I'd suggest it's (installer complexity) even less of a distinction now.

Aside - Microsoft installers always had it much easier than most GNU/Linux installers simply because Microsoft installers (for a very, very long time) didn't give a flying if there was an existing operating system on the box. They'd just re-partition, re-format, install, and overwrite the MBR.

Pretty much every GNU/Linux installation process took great care to accommodate other OS's already in situ, and also documented the process and correct ordering required to be sociable with other OS's.