Until you spend a lot of time in technical tasks, you might not be aware of just how significantly different the output from different models can be for coding and other technical tasks where we want discrete results and not wordy soliloquies. After weeks of deep diving into using several models for technical research GPT-4.1 is a clear winner by quite a stretch. Keywords has a great breakdown of the two latest ChatGPT models, 4.1 and 4.5 and how they stack up (and why) for different tasks.

https://www.keywordsai.co/blog/gpt-41-vs-gpt-45-a-comprehensive-comparison

Refactoring huge amounts of code. It's been quiet around here.

@Oksana the two hypervisors no one should consider.

@EddieJennings said in Ubuntu Server "Search Domains":

@scottalanmiller said in Ubuntu Server "Search Domains":

It's not a LInux thing, it's a general networking thing so the same on desktops, Windows, wherever.

This is correct. If I recall right, within the Windows network config GUI, it's called a "domain suffix.".

Sounds right.

If you use Active Directory on Windows, it fills it at least one search domain automatically. Windows is about the only place to see search domains still used frequently. This is primarily used for when Linux is put onto legacy style Windows based LAN networks.

@CCWTech they serve little purpose in the modern world. It's just the network domains that are automatically tried when you put in a URL with only the hostname and leave off the network name. Which no one does, so it's unused. It's not a LInux thing, it's a general networking thing so the same on desktops, Windows, wherever. Just a standard part of DNS, but like many DNS functions, one that is mostly a vestige.

Just got this press release..

On July 2, 2025, Juniper Networks was acquired by Hewlett Packard Enterprise (HPE). As this integration moves forward, we will continue to steward your data with HPE's help. The personal data we hold about you is being transferred to HPE and will be processed by them in accordance with their privacy statement.

If you currently receive marketing emails from us, you will continue to do so, and your preferences will follow your data to HPE. You can manage your preferences using the Juniper preference center.

Yours sincerely,
Juniper Networks, a Hewlett Packard Enterprise company

@CloudKnight said in Rethinking Virtualization? Start with Hyper-V 2025:

I wouldn't even consider hyper v. Proxmox and xcp-ng are the future

And like, they were the "now and future" like, a decade ago. This isn't a recent thing.

@EddieJennings yeah, that would be nice for sure.

@Oksana LOL, if you are considering Hyper-V in 2025, you should be looking for a different career path.

@Oksana VMware still exists? LOL Why?

@CCWTech Looking at it, I think that it requires a client. Which isn't all bad, but limiting. The convenient web interface is paid only.

While generally not recommended, sometimes we need to back up to a USB drive or other hard mounted backup location in ProxMox.

Problem: USB and other external devices can easily lose their mount while a system is already online causing the mount point to unmount forcefully and leaving a folder where the mountpoint had been. Rarely do we want the backup function of ProxMox to backup to that local folder, we rather want the backup mechanism to fail and then have an opportunity to alert us to fix the backup location issue.

Solution: Inform ProxMox at an application level that the backup location is a mountpoint and not a local folder so that it runs detection prior to backing up.

What to do?

1. Get your list of storage locations from ProxMox
   pvesm status

2. Find your backup location NAME on that list. It is a ProxMox label, not the folder name. For many of you it might just be called something like "backup"

3. Inform ProxMox to test if the location is mounted prior to backing up. Replace the word backup in this example with the NAME you got from step 2.
   pvesm set backup --is_mountpoint yes

4. You are done if that succeeds silently, but you can double check that the change has been made by looking for it in this file:
   cat /etc/pve/storage.cfg

We have Tactical, it just automates MEshCentral poorly for Windows and doesn't work for non-Windows at all. We actually have to replace it because of that.

If you haven't played with it for tiny scale stuff, TailScale makes this SO easy to do.

@Mario-Jakovina you are correct, in bridge mode a telecom router SHOULD be essentially transparent. It's just an extra switch at that point.

MT is good. I always prefer a dedicated router/firewall. Then my APs separate.

If you run Ubuntu or other Linux system using the UFW firewall mechanism, and you probably want to limit at least some ports to only receiving traffic from CloudFlare's proxy servers. CloudFlare provides scripts for iptables, but not for UFW. But they do provide their IP list in a handy format. So here is all you need to do...

cd /tmp
wget https://www.cloudflare.com/ips-v4 -O ips-v4-$$.tmp
wget https://www.cloudflare.com/ips-v6 -O ips-v6-$$.tmp

for cfip in `cat ips-v4-$$.tmp`; do echo "ufw allow from $cfip to any port 80 proto tcp"; done
for cfip in `cat ips-v6-$$.tmp`; do echo "ufw allow from $cfip to any port 80 proto tcp"; done```

Notice this doesn't take action, it produces a handy human readable set of ufw instructions that you can audit before running. Just copy the output to a file and "bash file" to take action. Make sure to set the port to 443 in most cases, or run twice, once with 80 and once with 443.

Back in the day there was LogMeIn, but they lied about being free and it became very costly and difficult for people to get away from it. I've yet to encounter any competitor to MeshCentral that sits in the same space.

@Oksana Does it make it arbitrarily slow and buggy too? Does it break updates constantly for an "authentic' Windows experience?

If you need to remote into a Windows machine and get a list of printers without interrupting the user, this powershell command is quick and easy...

Get-Printer | Format-Table