On Sun, 8 Dec 2002, Lee Maschmeyer wrote: > Hello! > > Facing reality, folks: > > If you ask this group to compare Windows and Linux at any time for any > purpose in any version whatsoever, guess who's gonna win! > > This isn't surprising since the members of this group generally know Linux > better than Windows. People feel better about and are more comfortable with > things they know well. Moreover, since Linux is an inherently simpler system > and tries to do less for you, things are likely to be more straightforward. > > However: > > Uninstall _programs_ are _programs_. Just like any other programs they can > have bugs, and the emotional investment of programmers being what it is, > it's highly likely uninstall programs receive relatively little testing. > Further, anyone who doesn't know an operating system very well is likely to > attribute to it ills that people more familiar with it will wish to correct. > I, for example, know Windows much better than Linux at this point, so I may > have erred here. On the other hand, someone said people will replace Windows > libraries with their own versions because the standard ones are deemed not Here is a true story that illustrates the problem. It actually happened on OS/2, but it could equally-well happen on Windows. I had a system with a Diamond Stealth video card, and it came with OS/2 drivers and an install routine. In my case the install routine didn't actually do anything, but in researching it I discovered other versions for other Diamond Multimedia products did delightful things such as: Unconditionally patch the kernel. The patch was correct for one level of OS/2 2.1, but it would also apply the same patch to OS/2 3.0 where it was decidedly wrong. Unconditionally replace a kernel module. Again, it worked for one level of OS/2 but no other. This is the kind of thing I was thinking of for Windows. I used IBM's Visual Age C compiler, and if I distributed software, I was explicitly allowed to distribute some components of it. I was expected to rename those components, but if _I_ didn't and _you_ didn't, and Barbara installed your product and mine, and you and I distributed different versions of those shared libraries, guess what happens? Probably, one of our applications works (assuming it's bug free), and likely the other doesn't. On Linux, this doesn't happen. For starters, people simply do not distribute components of other packages with their product; at least, I've never seen it happen. They _may_ distribute the entire component as a convenience. If you are using a package manager such as rpm, then a) It will not replace files from another package unless you explicitly instruct it to do so; b) It will not install a package if the correct versions of packages it requires are not installed, unless you explicitly instruct it to do so. > Personally, I've always liked the Windows model of putting the whole > component (aside from shared system libraries) in one tree; executable, > libraries, help files, manuals, DLLs etc. are more than likely all in the > same place. Delete that tree, you delete everything. In Unix, though (and I OS/2 does that. The path was getting extraordinarily long when I switched to Linux, and long paths are bad for performance. Ditto libpath (used for DLLs). > assume Linux), you've got binaries under some flavor of /bin or /usr/bin or > /usr/local/bin, manuals under /usr/man or related (or unrelated) entities, > libraries under /lib or /lib/share or whatever ... An uninstall package that > remembers to grab all of these will uninstall everything; a package that > forgets something, or a user who forgets something, will leave stuff behind. > > Bottom line: The uninstall is as good as the uninstaller, not as good as the > operating system. Most Linux distributions use either rpm or dpkg. rpm and dpkg have a database of packages and the files they contain. That's how they can prevent new packages from overwriting old ones, and how the can remove complete packages correctly every time. There _are_ no separate uninstall programs because they're not needed. > > -- > > Lee Maschmeyer > lee_maschmeyer@wayne.edu > > "Now I've reached the awkward age of man: > Too old for Castoria and too young for Serutan." > --Homer & Jethro > > > > _______________________________________________ > > Blinux-list@redhat.com > https://listman.redhat.com/mailman/listinfo/blinux-list > _______________________________________________ Blinux-list@redhat.com https://listman.redhat.com/mailman/listinfo/blinux-list