[crux64] multilib vs. pure 64

Johannes Winkelmann jw at smts.ch
Thu Jun 26 08:25:33 UTC 2008


Hi there,

I'll start a separate thread to avoid polluting the "soft-dep" thread. A
number of people seem to have rather strong feelings against multilib,
however i get the impression that the impact is in fact not that big. To
help us with deciding, here's a summary of what I think would be a good
approach:

1. core-x86_64/{glibc,gcc,binutils,bin86} multilib
2. all the rest of core-x86_64 and opt-x86_64 is pure 64-bit
3. we offer separate compat32 repositories (as compat32.rsync.inactive),


For the pure 64 user, the only drawback is the additional 32-bit
binaries for glibc, gcc and binutils. At the same time, the user has the
option to add the 32-bit compatibility layer should the need arise later
on, without the need to replace glibc/gcc on the running system.
- there's no performance penalty
- there's no additional issues (since compat32 is purely optional) which
  can happen on such a system that wouldn't on a absolutely pure64 one 
- there are no workarounds/hack required in Pkgfiles, pkgutils or
  pkgmk.conf


For us developers, it means that with a single platform, we can cover
both groups of use(r)s. Those that care about compat32 can join as
maintainer in these repos as well, 
- there's no additional work involved, we have to maintain gcc et al
  anyway, and helping in compat32 is optional



So to me, it boils down to this:
- some disk space "wasted" when running only 64-bit binaries
+ chance to add the compat32 layer at any point without any hassle
+ single platform, which means all development efforts, bug solutions
  etc. help both the "pure" and "non-pure" users.

Anything I missed? 

Thanks for reading,
Johannes
-- 
Johannes Winkelmann              mailto:jw at smts.ch
Zurich, Switzerland              http://jw.smts.ch



More information about the crux64 mailing list