Hacker News Reader

Comment by pxeboot | Hacker News Reader

Very valid issues here.

Although it has improved, Windows Update is still shockingly bad compared to any decent package manager used on Linux in the past 10+ years.

View on HN

I'm perpetually shocked by how slow many of the loading / progress-bar screens are in Windows. Random operations taking 1-5 minutes without any visible change is routine.

Honestly, what the unholy fuck is actually happening behind these screens? The only thing that takes anywhere near that long on Linux is compiling.

Yes, one such bug was covered here:

- https://news.ycombinator.com/item?id=21743424

- https://randomascii.wordpress.com/2019/12/08/on2-again-now-i...

I never got into an intractable package issue with windows, where the dependencies were broke in a way where I had to crawl through logs and manually delete things to get them working. That’s happened to me a handful of times with Ubuntu and bad package configs.

> I never got into an intractable package issue with windows, where the dependencies were broke in a way where I had to crawl through logs and manually delete things to get them working.

I cannot count the number of times that's happened for me on Windows, Linux and macOS. None of them are immune to this problem.

I've had numerous unrecoverable blue screens after Windows Updates.

I think this is inherently due to Windows registry and no enforced standards for Windows file system. In Linux, packages are generally reviewed to ensure they follow a set of practices to ensure that they use the appropriate directories. Then they simply have to extract the files/folders to install the application, and removing is as simple as deleting those directories/files. In Windows, you really never know what you're going to get when you run an install. You might get 1k registry keys created, and it is near impossible to monitor and replicate. They might install the App in Program Files, or if it is new Appx package it will be somewhere completely different. Some apps may install in AppData. It is completely unpredictable. Running an uninstall may call msiexec, or some other uninstaller.

In order for Windows to have a Linux like package manager, they would have to re-define the install process completely to act like Linux and probably move away from the Windows Registry completely.

I'll take the Windows Registry over an OS that litters my home directory with dot files, or even simply force me to create my own format of saving my settings (even ~/.settings is barely an improvement).

Saving your settings should be an API offered by the OS and the Windows Registry is pretty good at that.

Have you looked in a Windows user's home directory?

What does this question have to do with the existence of the Windows Registry API?

Because you said that you’d take an OS with a registry over one with dotfiles, and his point is that these days you have both

The last Windows version where updates worked more or less OK was XP.

With Vista they introduced WinSxS and while in theory it was a great idea, it incurred some extremely heavy tradeoffs in terms of updates.

Recently I've upgraded 2 machines from 1903 to 1909 (aka the "semi-annual update"). Both times the process took less than 10 minutes.

Going from 1903 to 1909 is only a ~1mb enablement package [1]. It essentially changes the version number in the registry and enables several existing features.

[1] https://support.microsoft.com/en-us/help/4517245/feature-upd...

Is there anyone who can explain why Windows Update is so much slower than running »apt upgrade« and similar?

Because apt upgrade, or pacman -Syyu is simply extracting files and using hooks to generate a new initramfs. Windows Updates has to do a lot of operations to ensure the registry is correct, and then batch upgrades to create and remove scripts to run after restart.

Part of the reason is what I described in https://news.ycombinator.com/item?id=20812142 and its child comment.

Interesting! Thank you for writing this up.

I have wondered whether the servers are overloaded, will maybe try logging the network traffic the next time I try to update.

What is this comment supposed to mean? How are you comparing the performance of updates across OSs?

The relatively common (?) experience that:

1. Running `sudo apt update && sudo apt upgrade` is much, much faster than explicitly going to the windows update screen, checking for updates and then updating.

2. Linux OSs are usually stuck "updating" while shutting down and restarting, far less frequently and for a far shorter time (if at all), than Windows.

Obviously the comparison isn't entirely fair (though in some ways the comparison is biased against Linux, since its update also usually includes the vast majority of user-space programs), but comparing the percentage of "updating time" of two OSs (whatever the technical reasons) seems to be reasonable, to me.

Yes, this is a very good explanation.

Do you have any numbers for this? It doesn't match my experiences at all.

Edit: it's been my experience that most people who think Windows sucks think this because they use it exclusively at work where their IT department has gone mad with power and turned off all the good stuff, and the network is slow.

> Do you have any numbers for this? It doesn't match my experiences at all.

Sorry, no, purely "anecdata", mainly based on a variety of personal and family laptops and desktop computers. There have been occasional IT department-managed update nightmares, but I don't count these.

Just for the record, I didn’t say that Windows suck :-)◊

◊ I mean of course it does since every OS sucks, but that’s a given.

There was a promise to make this magical and better apparently. I haven’t seen an improvement yet.

Edit: to the downvotes, sorry for being disappointed in having windows Update ram a whole core at 100% and eat up my AWS CPU credits to do precisely SFA.

Reply on news.ycombinator.com