[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: relocatable packages ?

On Wednesday 22 June 2005 06:35, Jason L Tibbitts III wrote:
> >>>>> "sv" == seth vidal <skvidal phy duke edu> writes:
> sv> Well if you read the fhs there is a good reason to keep matlab out
> sv> of /usr.
> sv> http://www.pathname.com/fhs/pub/fhs-2.3.html#THEUSRHIERARCHY
> sv> read the 4th sentence
> I don't put anything in a direct subdirectory of /usr.  For me,
> Matlab mostly lives in /usr/lib/matlab-14.2.
> There is no definition of "large software package" on the page you
> link; I think Octave, Mozilla and R would qualify, and they all live
> entirely under /usr.
> I would put Matlab, Mathematica, Splus, the PGI compilers, and other
> such things in /opt if they weren't integrated into my distro.  But
> they are essentially indistinguishable from any other package; if I
> enter "yum install mathematica" I get Mathematica.  It's as much a
> part of my distro as any core package, anything from extras or
> another repository, or any package that I built an RPM for.

Where does your Mathematica rpm come from?

I haven't looked at R or Octave.  But I think the placement of Mozilla 
into /usr/lib(64)/mozilla* probably makes sense given the structure of 
the Mozilla tree and its purpose.

Matlab, on the other hand, has a totally non-fhs directory structure, 
and given that e.g. users are expected to run executables within the 
Matlab tree (from etc/ as well as from bin/), it doesn't make sense to 
simply plunk the whole matlab tree down 
into /usr/lib/matlab.  /opt/matlab, on the other hand, seems fine to me 
-- the point of /opt is to shrug and take whatever each vendor package 
throws at you.

You could write a matlab specfile that does a lot of careful work to 
move things around from within the Matlab distribution tree into 
appropriate places in /usr.  You might find out, though, that that 
breaks matlab, or at least requires a fair amount of patching matlab 
files so they know where to look for other pieces of matlab.

Whatever the wording of the FHS documents, I don't think the choice of 
what goes in /opt has so much to do with size as with lack of fit in 
the /usr heirarchy.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]