Re: Interface Stability in GNOME

[Ed Hunter <ed hunter Sun COM>]

Havoc is right on the mark in the sense that the stability levels are used to 
communicate an implied contract between Sun and the consumers of the interfaces 
we produce.  It seems like there's general agreement that there is at least a 
contract between the GNOME community and the ISV's.  There may be other 
potential contracts between GNOME and other groups (developers?) but let's just 
focus on the ISV case for now.  

To a certain extent the community needs to decide what level of flexibility they 
want to allow with the ISV's.  Havoc's assertion is that there are two degrees 
of freedom, use it or don't.  Others on the thread might assert a third degree, 
use at your own risk (unstable).  There might be others.  

Within the context of those degrees of freedom you need to decide what the 
contracts will be and how you'll enforce them.  For instance, you all could 
decide that stable means "could change once a year" (not a great choice given 
the word but possible).  You could enforce that by some form of testing or 
examination of changes to the source.

Part of what the Sun process does is try to be clear about what the expectations 
are for each of the stability levels.  Particularly around what types of change 
(and how often) the consumer can expect.  As Brian pointed out inside the 
company we track a bunch of other things that a mainly interesting because of 
changing organizational structure.  For instance a contracted private interface 
in theory will survive various reorgs since the group acquiring the technology 
would be bound by the contracts (or at least know what they're getting 
themselves in for).  It also ensures that when changes get made people aren't 
suprised by them and that the provider of a "private" interface knows who their 
customers are.

So, it seems like there is at least agreement on having a stable and private 
bucket.  There's discussion on the use of unstable and potential discussion on 
the addition of other things (internal).  Does that reflect what people are 
thinking at this point?
-edh

> 
> On Thu, 2004-12-23 at 16:37 +0000, Ghee Teo wrote:
> >     If this is the reality of the situations they should be classified 
> > as Unstable even
> >     though they are on the developer platform. The ISVs  use these at 
> > their own risks.
> >     There is no point in hiding the risks from t
> > hem where trusts need to 
> > be built.
> 
> I certainly agree, for ISVs there are only two categories, use it or
> don't use it...
> 
> Havoc
> 
>