Quoth Walter Tautz, > I would slightly concur with your assessment of dselect but > only in the sense that it lists simply too many packages. I > am not sure, but perhaps it is possible to present the various > software types in a less verbose manner. Sort of like the slrn > news reader where one could collapse dependency trees depending > on the software. It would be nice to collapse certain dependency > trees while maintaining others.. Every now and then on the list we have a healthy flamewar about the pros and cons of dselect. Those who are new to debian argue that dselect is unintuitive and hard to use. Those who have an emotional investment in dselect (ie., they wrote it, or experienced the pain of learning how to use it), argue that online help is available and easy to find. While online help is certainly available, that doesn't cut it if we want debian to be a usable distribution. Yes, Unix/Linux is hard to learn, as it should be, but many people will be put off if they can't make it through such an essential step in installation. The fact that so many people complain about dselect is suficient proof that it's too difficult to use. The fact that so many people accidently delete all the packages in their system establishes that it's not designed for novice users (ie., those installing debian for the first time). Part of the problem is the wonderful number of packages available to be installed in debian. While 4000+ packages is great for debian users, going through them all in the one list in dselect is clearly unreasonable. Maybe collapsing categories or something would be useful, but it's unworkable as it stands. There are, of course, other alternatives to dselect, such as aptitute and capt, but AFAIK neither of these are installed by default. With the tasks packages it may not be as necessary for those installing to even see dselect, but if you want to browse the available packages, it's probably more likely to be used than: cat /var/state/apt/lists/* | grep ^Package: |less or similar. Unfortunatly dselect is something of a sacred cow in debian. It's like an ugly and dangerous, but historically significant, building that can't be renovated or pulled down, but which people try to avoid looking at. Maybe one day it'll be replaced by something easier to use, but I'm not holding my breath. I'm sure people are going to disagree with my assessment, claiming that dselect is user friendly, it's just that all the users are too stupid to use it. So maybe we need some good data. Is there anyone out there in debian-user land who is into computers, studying psychology[1], and looking for a little project? Set up a little experiment getting niave (windows?) users to try and install a number of packages using dselect. To make it interesting, throw in a few examples with hairy dependencies. Keep track of things like number of errors, time taken, use of help, and the number who result in a totally fscked system. There is shite loads of literature out there on HCI and usability studies, so you should have no trouble setting it all up properly. Feed back your results to debain-user, and to the maintainers of dselect, capt, aptitude, etc. This is something I'd love to do myself, but unfortunately I've now moved out of psych to criminology (while HCI is interesting, homicide is much cooler!). Cheers, damon (who has his asbestos suit standing by...) [1] HCI and usability isn't something that should be done by computer scientists unless they have a particular intrest in the field - they bring too many prior preconceptions to the studies. -- Damon Muller | Did a large procession wave their torches Criminologist/Linux Geek | As my head fell in the basket, http://killfilter.com | And was everybody dancing on the casket... PGP (GnuPG): A136E829 | - TBMG, "Dead"
Attachment:
pgpVUOf90Duwc.pgp
Description: PGP signature