On Fri, 2004-06-04 at 11:31, Aaron Bennett wrote: > ok then, according to the spec all of the Fedora.us (Fedora Extras) > stuff should go under /opt. Sorry,but this is such a straw man argument! The Fedora Extras (and other repos as well) packages are made specifically to fit in with the Fedora distribution and can thus easily be seen as system-native packages that should go in the /usr hierarchy. And really, that's the obvious, sane reading: packages that are built for the distribution's native packaging system should go in the /usr hierarchy. "Add-on packages" would typically be (often binary-only) programs that are not available packaged and that the sysadmin doesn't want to spend time repackaging so that it integrates nicely into the distribution package system. For example, I have MATLAB on a few machines; I'm mot going to spend time packaging it up. Thus I dump it in /opt/matlab, knowing that /opt is a nice place to dump stuff that I need to take care of myself and that I don't have to worry about interfering with RPM when I do that. > Luciano Miguel Ferreira Rocha wrote: > > Sun's java rpm doesn't in my linux system. > > > > All of their Solaris stuff does under Solaris... that's what I meant. Come on. Sun being boneheaded is completely irrelevant to Fedora. > > Why should any package from any vendor for any system that allows seamless > > installation, removal, listing and verification have to go to /opt? > > > > I don't know. I'm not advocating moving anything to /opt. For a number > of reasons I hate it. What I'm saying is just moving a bunch of stuff > to /svr won't achieve FHS compliance unless /opt is addressed. Only with some stupid fundamentalist reading of the FHS. Read what it says but no more. The most reasonable reading is that "add-on packages" are programs that are not integrated into the distribution package management system. > I think > that the FHS is a bunch of hooey, to be honest. It seems extremely sane to have a consistent definition of where things should go on a Linux system so that one can handle several distributions without too many headaches (and in some cases make packages that can work on several distributions). FHS might not be perfect, but it's there for a very good reason, and adhering to a slightly imperfect standard (although not to some insane obscure fundamentalist (mis)reading of it) instead of increasing fragmentation in the name of perfection is a very pragmatic and good thing to do. /Per -- Per Bjornsson <perbj@xxxxxxxxxxxx> Ph.D. Candidate, Department of Applied Physics, Stanford University