Hacker News Reader

Comment by _ea1k | Hacker News Reader

To counter your anecdotes, here are some more anecdotes:

- Windows XP - Power went out at a critical stage of updates. Completely hosed the system, and wasted hours of my life having to rebuild (including rebooting multiple times to get updates installed again after the recovery).

- Windows 7 - Cable giving my router IP addresses. Called ISP tech support... they had me manually plug the cable in and type cryptic commands at a DOS prompt. Is that unacceptible too?

Also, I'm really baffled as to how the network going down could kill a Fedora update. The downloads happen before any significant filesystem modification in every version that I have ever used (certainly any recent one). I guess maybe some third party package does something weird during a POSTINSTALL script? I dunno.

View on HN

The problem here is that the super poster is criticizing the experience under Linux when you install any of the OSes on non-vetted hardware, whereas 99% of Windows computers come preinstalled with all the drivers properly configured.

If your Windows system breaks, and that means anything beyond "oh I'm missing a driver", you have no recourse, you have to reformat. You end up googling and doing the exact same procedures you would under Linux, except because the OS internals are obscured and proprietary, you don't have whole stack detailed analysis on whatever problem you have, you have a black box and you have to hope the tools MS gives you are good enough.

It doesn't change that for 99% of users, regardless of OS, when you encounter a problem you don't try to fix it yourself. Most people aren't going to even try, they are either going to take the system to an expert who would use a terminal no problem or just get another one.

I experienced this myself recently. The Netflix metro app decided that something was wrong with the DRM path on my two-day old install of Windows 8.1. After much googling, I ended up having to re-install the OS. It was a very similar experience to when something breaks on a Linux install, except that I had less ability to go digging into the internals (not that that would be helpful for someone who doesn't know a lot about computers).

I think the primary difference with Linux is that developers are fairly content mucking about with things, so breakage that would be a showstopper for someone who is not intimately familiar with computers goes somewhat unnoticed. It's hard to step back and say "what would I do if I didn't know about <x,y,z>?" in daily life.

I've been intentionally writing KCM control modules for various traditional command line functionality recently, with the intent to get them in some early series 5 KDE release. Just proof of concepts, but so far I've gotten working ones for xrandr, wlrandr, and saned (ie, network scanners). They aren't usually very complicated, just a bunch of widgets. I intend to finish them with a qml rewrite, most of the work was before you could embed qml in widgets.

The point is, yeah, you shouldn't have to go to a terminal, but that requires the development of guis to overlay the complexity. I use my relatives as test beds for whatever breaks that requires a "control panel entry" to fix.

Yeah, hasn't Fedora used transactional updates for ages now?

Power outage is a harder ball.

Reply on news.ycombinator.com