> All in all, there just isn't a lot of advantage to moving to a different > packaging system. There could be slightly more software available, but it > would require maintaining two systems in parallel and reformulating all of > the tools.
I completely agree. You shouldn't move to another package database system. The idea that the file system is the package database is both simple and versatile. What I do hope is that you will consider using another method for software installation that would interact with your package database. Zero Install is a way of updating software, doing various security checks, adding/removing menu entries however the placement of software is flexible enough that it could be used with your file system hierarchy and probably your current packages. What I would recommend is that you would use your current method for distribution upgrades, and use Zero Install to do all other installations as it provides non-root installation. In order to maintain your own database, I would recommend that all users can edit the /Programs folder, however they cannot edit files or folders that were created by the root account (like Xorg-Server). > And we generally shy away from encouraging people to install from > arbitrary locations on the internet The fact that users can easily install software from anywhere is, I believe, the single most important reason why Windows and Mac are much more popular than Linux on the desktop. Various issues with third-party software installation are covered at Ian Murdock's blog (elected chair of the LSB, and creator of Debian), Liquidat's blog and the Autopackage documentation if you're interested. http://ianmurdock.com/2006/12/15/software-installation-on-linux-today-it-sucks-part-1/ http://liquidat.wordpress.com/linux-software-installation/ http://www.autopackage.org/docs/devguide/autopackage.html?q=hosted#binport > I've always supported alternate package schemes under GoboLinux. I > did some investigation into Zero Install many years ago. It should be > easy to integrate today. The biggest limitation then was the lack of > software available. I tried Autopackage as well and suffered the same > problem. I had better luck with Klik. I have about 70% success rate. > My biggest limitation then was my sat internet connection and klik > didn't like the latency. I agree the lack of packages available for Zero Install and Autopackage is not optimal however the co-operation between Autopackage, Zero Install and hopefully GoboLinux would mean more packages could be available for all of them. In my personal opinion, Klik is sub-optimal at best. >>> the binary must be portable >> This is *not* a requirement, and in general packages will not work if you >> move >> them elsewhere. Sometimes they do, but it's not planned that way. I apologize. Thank you for correcting my error. > I think these generalized package tools would really benefit from an > enchanced package selection. I would encourage they look at Compile. > They would benefit from a tool that is truely prefix agnostic. Hm, this is an interesting idea. I shall talk to the Zero Install developers. > If they can develop a community around a tool then the distros would > benefit from adding support out of the box. That is how the real > winner will emerge. I agree that a community is required, else a project will fail and ultimately become useless and unused. However, I think the first response I got to this suggestion will be typical of many distribution vendors. Their investment in their package management system is so great that they refuse to consider alternatives, even if an alternative could offer integration and a significant feature set to their distribution. And, I believe the hope that Linux distributions shall all conform to the one method of installation is futile because of the versatility of Linux which is both it's strength and weakness. You may question why I even bother to ask in this case. The answer is simple. I have hope regardless. ;) > Then commercial software will follow. This is an important next step > for Linux. I hope GoboLinux can help. I assume you mean proprietary by commercial (correct me if I'm wrong). In fact proprietary software for Linux is alive, but certainly not the norm. Imagine being a third party software vendor looking into providing software for Linux. When you learn about the many package formats you are likely to require providing in order to be accepted, you would soon be deterred. This is the main reason why I consider the united effort between Zero Install and Autopackage significant. I also hope that you will take part. Thanks for your time, Matt _______________________________________________ gobolinux-devel mailing list gobolinux-devel@lists.gobolinux.org http://lists.gobolinux.org/mailman/listinfo/gobolinux-devel
