Good bye

Matthew Saltzman mjs at CLEMSON.EDU
Fri Feb 1 19:29:41 UTC 2008

On Fri, 2008-02-01 at 10:09 +0000, Stuart Sears wrote:
> Les Mikesell wrote:
> [ cut discussion of free vs opensource etc]
> > People should be free to make that decision for themselves separately 
> > for every component.  That's what standard interfaces are about, to give 
> > you that freedom.  There is nothing free about things that take that 
> > choice away by restrictions or refusal to provide usable interfaces for 
> > interoperability.
> I think you are making an argument that you will never win here.
> Fedora makes no guarantee of a stable API or ABI. This would be 
> impossible to do while trying to follow upstream - rebasing kernel 
> versions and so on is definitely going to interfere with this.

The old kernel development model was much more third-party developer
friendly in this sense.  It used to be that even minor version numbers
(2.4.x) were stable in this sense and odd minor numbers (2.5.x) were
not.  Then a distro could include the even ones and developers could
rely on the API and ABI until the next even release came out.  In the
meantime, they had time to adapt to changes being made in the odd
releases.  The odd releases were generally made available by distros for
testing by the stout of heart.

> If (for example) VMWare only works with kernel version <blah> and they 
> don't keep up with newer releases, then it will break if you update your 
> kernel. So don't (unless you know they have patched vmware to cope).
> There will be other software that falls into this category.
> Vendors do not on the whole support fedora because of issues like this.

Keeping up with rapid API/ABI version changes is a burden for any
third-party developer, not just proprietary ones.  Having to go back and
continually modify older code for compatibility distracts from
development of new features.  It also introduces these delays where
users can't move to new kernels (which might include critical security
fixes) or must forgo their third-party codes until their vendors catch

> If you want ABI/API stability to be guaranteed for the lifetime of a 
> product, then RHEL (or CENTOS or one of the other rebuilds) would 
> provide you with this, and I am fairly sure that the 3rd-party stuff 
> would work for them (I haven't tested).

Les's argument here is that it's hard to get the latest application
software without committing to live with the instability at the kernel
level.  RHEL5 is still shipping Firefox 1.1.5 AFAIK, and equally ancient
versions of other applications.  That may be fine for enterprise
managers, but many single users and developers would like to be able to
keep up with application and toolchain advances.  

People talk about TCO and point out that there's lots more application
software bundled with with a typical Linux distro than with Windows.
That's all very good, but unbundled applications have the advantage that
the user isn't tied to the bundled release of the application if they
want to stay with a particular version of the system.

> Fedora is a moving target, which is something that its users have to 
> accept. Most of us do.

Some of us accept it--or even welcome it--for some things but find that
it's a burden for others.  There's no reason not to consider changes in
policy that would address some of the disadvantages of this situation
without doing too much violence to the advantages.

> [snip more of the same argument, more or less]
> oh, as for not providing immediate access to closed-source or in some 
> cases US-patent-infringing software, this is also part of the fedora 
> 'mission' and is also a safety issue.
> Fedora receives a fair amount of backing from Red Hat, who have some 
> money and are based in the US - they become an obvious target for 
> money-grubbing lawsuits, which I suspect would not upset some other 
> large software companies. If you don't like this, you are free to respin 
> a Fedora clone of your own that does provide this access out of the box.
> many people seem to understand why Fedora has this restriction. It's not 
> likely to change.

I don't mind dealing with the need to go get third-party software from
third parties.  One has to do so for Windows as well.  The set of things
that one needs to get from third parties in order to have a system that
performs most of the common tasks a user expects is just different.

But there seems to be an attitude on the part of some people in the
community that the best way to pressure vendors of proprietary software
to open their code is to force users who need that software to suffer
without it.  I think that alienates users and is counterproductive.

> Stuart
                Matthew Saltzman

Clemson University Mathematical Sciences
mjs AT clemson DOT edu

More information about the fedora-list mailing list