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

Re: RFC: Better portability for package maintainers

On Sun, 2006-05-21 at 09:49 +0300, George Danchev wrote:
> On Sunday 21 May 2006 05:35, Erast Benson wrote:
> > On Sat, 2006-05-20 at 21:11 +0200, Steinar H. Gunderson wrote:
> > > On Sat, May 20, 2006 at 11:51:09AM -0700, Erast Benson wrote:
> > > > Do you really believe so? Do you understand that such a "hybrid" will
> > > > not run any existing Solaris apps like you will not be able to run
> > > > simple thinks like Macromedia flush player, JRE, JDK, Oracle, SAP, etc
> > > > etc... Do you still wanna do that?
> > >
> > > Erm.
> > >
> > > If Oracle and SAP are on your list of “simple things”, what then are
> > > large complex things for you?
> >
> > But I hope you still got me right. For me, all these "things" are
> > existing applications which must run. The world is not 100% open sourced
> > yet and we are in it, we are part of it, therefore my ideal OS need to
> > be capable to run existing freeware and closed binaries as is without
> > re-compilation. I want to run VMware, Oracle, Skype, SAP, Macromedia
> > flush, etc, etc, etc. I want my Nexenta to run DTrace, BrandZ
> > virtualization, ZFS, Zones without major re-design, etc, etc, etc...
> >
> > Once you accompany OpenSolaris kernel with GLIBC, you will kill this
> > capability, you will not be able to run anything other than OSS compiled
> > for your particular distro. That was my point. And isn't LSB is what
> > GNU/Linux moving towards to? In OpenSolaris we have its Core which we
> > following as a standard and I don't see any single reason not to do so.
> You have your points right, but you should realize that Debian GNU / <Kernel>, 
> is glibc based. This means that your Base System without the kernel should 
> come from GNU sources. Having that said, you should invest some efforts to 
> port glibc to the Solaris (or OpenSolaris, Nevada, whatever[1]) kernel (to 
> support all these fancy features mentioned above), as this has been done for 
> glibc and the FreeBSD kernel by Bruno Haible.

I'm personally will not do that. As I said earlier, I did it a year ago,
I even managed to run statically linked binaries on GLIBC + OpenSolaris
kernel. Than I realized that the resulted Operating Environment will not
be compatible with *anything* existing... how much it will be better
than GNU/Linux or GNU/OpenSolaris or SUN/OpenSolaris? I realized that
porting effort might be greatly minimized by utilizing different

1) provide 100% Debian environment, so native Debian scripts will run as
2) extend SUN C library with missing GLIBC functionality;
3) use of side libraries like libiconv, gettext, libintl;
4) use of transitional packages.

As the result, we are fast approaching to the point when all existing
Debian APT repo will be fully ported to Nexenta. We have 7000+ packages
at the moment and will probably have 10000+ by the end of next month.

> On the other hand if you go for Solaris [1] own kernel and libc and porting 
> Free Software on the top of that, your Nexenta OS is as much GNU as say MS 
> Windows or Mac OS X since such (non-core or non-base) applications could be 
> ported and compiled on them too, but these OS'es does not have GNU in their 
> names (yet ;). Thus in this case Nexenta or GNUSolaris should be named like 
> Nexenta Sun / OpenSolaris ( as in Distributor Base / Kernel ;-)
> P.S. no offence implied, just sharing some thoughts ;-)
> [1] http://opensolaris.org/os/community/onnv/
> Note that Opensolaris / Nevada are not GCC ready yet, Sun Studio 10 is the 
> preferred compiler
> -- 
> pub 4096R/0E4BD0AB 2003-03-18 <people.fccf.net/danchev/key pgp.mit.edu>
> fingerprint 1AE7 7C66 0A26 5BFF DF22 5D55 1C57 0C89 0E4B D0AB 

Reply to: