[oi-dev] OI project reboot required

[Jim Klimov]

On 2013-05-09 16:02, Bob Friesenhahn wrote:
> The Tribblix approach is likely a good one.  Start off with a good
> smaller core and then add more sophisticated features via packages.
> This requires a new distribution though.


Two words: "backwards compatibility" ;)

Reinventing the wheel from scratch doesn't mean that it magically
becomes incompatible and should be renamed and called a new project.
If it is round, and weight can be put on an axle, it is still a wheel.

As long as the administrative (end-user) interfaces remain in place,
the ways we get there can change. In case of OI, as a multipurpose
distribution which is an upgrade path for IPS-based OpenSolaris (and
older OI) systems with SVR4 support, there is just a certain set of
properties that should remain in place. Whatever the process is under
the hood - we don't really care, all it should do is create the IPS
packages and spit them out via pkg.oi.o repository server.

I am not convinced that a new system rewritten from scratch won't be
able to make an output indistinguishable (to a reasonable compatible
extent) from whatever the current "ugly difficult" process produces.

In short, change of the process does not require that this becomes a
new distro. Especially since the impact on deployments would be minimal:
there's just a few people that know the current process and practice
its black magic, and are allegedly fed up with doing so and have little
personal attachment to have that remain in place. If a functionally
equivalent compilation technique appears, which a wider audience can
use, there are nearly zero systems that must be "converted" to it and
won't know that they should do so (i.e. flag-day changes). Likely,
those systems that fulfill this role now would be the ones that will
implement the new method first - if not develop it directly ;)

//Jim