Just a minor hiccup. I recently installed a kernel update on my laptop running Debian 10 (buster) and, for whatever reason, it didn't install the Linux kernel headers package. A couple of days later, when attempting to launch VMware Workstation, it would fail showing an error that it couldn't find the headers for the running kernel. I installed the package, launched the application once again, and now it showed a different error about a kernel headers mismatch. The solution is published on the VMware Knowledge Base here. Just run the following command as root:

# vmware-modconfig --console --install-all

That will reconfigure things with the new kernel headers, and will allow you to launch the application. {link to this entry}

I was recently called to assist with a server that was experiencing problems to boot and, in the process of troubleshooting the issue, noticed certain messages in the syslog indicating that systemd had unmounted the /boot/efi filesystem. To the surprise of everyone involved in the call, that is a "feature" of systemd. In other words, it can unmount a filesystem that the sysadmin specifically wanted to mount via fstab, not automount. I'm not sure I fully understand the reasons why it might do that. After all, if the sysadmin wants to use automount, he/she can do that. Who is systemd to make this sort of decisions? Worse yet, who is it to make it in a relatively silent manner? As it happens, plenty of other people are quite angry about it. See, for example, systemd bug #1378.

So, that incident got me thinking about the current status of systemd. Sure, it is so widespread these days that we have gotten used to it. And yes, it does have its advantages over the old System V init. However, the problem is not only that systemd was already a bit of a monster when it was first released (which caused plenty of controversy at the time), but the monster has been getting bigger and bigger. It continues extending its tentacles to areas that one just cannot see as legitimately belonging to an init system anymore. For instance, notice this recent article about systemd-homed. Apparently, this code has now been merged into the main systemd source tree to "make it easier to support migratable home directories, more self containment within home directories, better password encryption and encryption handling, and other modern Linux home directory features." Yes, it all sounds really cool. But do I really want systemd doing this too? Heck, since we are talking about systemd features, we can also add systemd-boot to the list, which appears to be designed to replace the boot manager altogether. What's next? Replacing the kernel itself? Yes, I know. I'm being facetious. However, I find it truly worrisome that a key component of the Linux infrastructure is becoming this bloated and, worse yet, we, the users, have little choice in the matter. As far as I know, every single major distribution has adopted it by default, and users have no choice to run something else. The most recent adopter of systemd as the standard init system was Debian, the distro I use, very well known for its stability, but also for providing the users with a very wide range of choices. And yes, I fully understand why they made this decision. And yes, it is perfectly reasonable from their point of view and, if I were one of their developers, would probably have made the same decision. Yet, none of that changes the fact that a very bloated piece of software that contradicts the main tenets of the well established UNIX philosophy has quickly become the de facto standard in Linux and we, users, have no choice in the matter.Sure, there are some minor distributions that, Asterix and Obelix style, resist the trend. However, they remain extremely unknown, small and, therefore, unreliable. And no, this is not the same as the current situation with well known desktop environments. At least in that case end users have a choice to run something else. My feeling is that the day will come when we will regret all this. {link to this entry}

A few days, while reviewing the latest technology news, I came across an interesting article by Robn Enderle titled Rethinking the PC: Why virtual machines should replace operating systems. To be clear, I'm not sure the title correctly reflects what the author argues. He is not so much arguing in favor of doing away with operating systems, but rather with the traditional idea of running a single operating system on each system. Yes, sure, that already happens on servers everywhere. But Enderle goes beyond that. What he proposes is to run virtual machines on personal computers, including laptops and home desktops.

Much of the problems we’ve had with deployments over the last couple of decades have revolved around the need for IT to keep the PC image static while the hardware changed. If we instead preloaded a Virtual Machine from either VMware or Microsoft – and then placed the image on that – you could assure a level of compatibility you generally don’t get today. (...)

On a PC, you could have distinct VM instances for work, school, and personal use with differing levels of user freedom. The company VM would be locked down so that the firm is better protected from the other usage models. Viruses often come into companies carried by employees who aren’t careful with their personal use of their firm’s PC. You mostly see this kind of behavior with developers who need to keep their dev projects separate from their enterprise image.

Even with a three-image installation (work, school, personal), you’d be able to optimize on all three support organizations. Work IT handles the work image, school IT handles the school image, and the OEM helps with the personal image (which they could charge for). You’d get a higher level of security because the two or three usage models would be isolated from each other – and you’d free up the OS vendor and the platform vendor to advance their platforms faster because they could specify a set Virtual Machine configuration.

The idea sounds quite intriguing, I think. I seriously doubt it will make it anywhere, but it's interesting. {link to this entry}