I thought of this interesting question the other day while messing around with Slackware 9.0 which was one of the last versions of Slackware to come on a single disk. The goal was to try to take a Slackware 9.0 install to the most recent stable and it was almost accomplished. Glibc was the largest hassle…and I made it to Slackware 11.0 before something caused things to not boot at all. All things considered, I spent 3 days on trying to get Slackware 9 to current.
Slackware for those of you that don’t know, has no dependency resolving package manager. Previously, a good attempt was made with swaret and that was my first jump into package managers with dependency resolution all together when it came out…but Swaret is no longer being maintained and doesn’t really work well anymore.
Since Slackware has no real dep resolving package manager…it’s one of the last ‘true’ Unix like Linux versions out there. Back in the early to mid nineties…things were exactly like this. If you wanted to update your Linux version…you stepped through it manually and tried to get things to work. What was great about Slackware was making your own Slack packages with source…no dependency resolution but in the process of making the package you’d have all the dependencies eventually installed. In this entire process, you became VERY familiar with your system…how it booted, what run level things occurred at, how cron jobs worked, etc. You were baptized by fire so to speak…you were to sink or swim.
As I said, this got me thinking…do we rely on dependency resolving package managers TOO much? They’re clichรฉ now of course…run of the mill. Back in the 1990’s though rpm was the only true package management system around…and rpm was never designed for internet consumption. The guys who wrote rpm had in mind CD and floppy upgrades. Fast forward to now and we have zypper, pacman, urpmi, deb, and conary…all built with online repositories in mind. Do these managers take the heavy lifting away for new users? Do they spoil them?
Do systems break less with easier resolutions due to package managers? Does it mean that the new user of today won’t be as experienced as the old user of yesterday?
I think it might.
Users in the past had to chip away and reassemble with less documentation and no package manager. This meant that the user of yesterday ripped apart systems and packages to discover how they worked and which cogs fit where.
The user of today follows step by step instructions and the software is given a sane set of defaults by most package developers when said package is installed.
Does this make for lazy users?
I don’t think users are lazy per se…but as previously stated, spoiled ones. And it’s no fault of their own…it’s the direction the software has taken us. Now the questions we need to answer are:
- Is this direction the correct direction we should be heading?
- Are there better approaches to package management that don’t follow the model we have currently (other than Conary)
- Can we come up with a system that doesn’t make new users spoiled?
I think I’m of both worlds…I started off with no package manager but managed to ride the wave of Red Hat 7.2 and above followed by Mand{rake,riva} and PCLinuxOS. I’m both spoiled and unspoiled. I know what it takes to manage a system without a conventional package manager but I also know how much time it can save me to use one. I sometimes find myself wanting less though…less and more. Less time and more hands on gutting the system. I think I’m in the minority though.
How about you, as a reader of this article? Do you think new users are spoiled by conventional package management systems? Do you see solutions or have ideas we can discuss? Is this really just a process we can improve or is there any programming to be done? Please sound off in the comments section!