[Proj] ICSM (Australia) transformation file licensing

[Martin Desruisseaux]

Le 18/01/2017 à 21:32, Greg Troxel a écrit :

> Everyone in the community knows that randomly changing projection code
> files to return wrong data is not helpful and I really don't think
> people are going to do it.
>
But isn't what Proj.4 did when changing axis order? I see units of
measurement issues (e.g. in WKT) as another example.


> A license that doesn't permit modifications is not compatible with all
> Free Software. It's not just GPL; such a license is also not
> compatible with the BSD license.
>
Indeed. But some issues are more important than compatibility with open
source licenses, and standard integrity may be considered as one of
them. Projects who change the data do that because they think that their
changes are legitimate, not because they want to be evil. A typical
example is ignoring some aspects that a developer considers as useless
complexity (e.g. axis order and units of measurement). It is okay to not
implement every aspects of a standard, but when ignoring an aspect
breaks the interoperability goal of that standard, we have a problem.
The EPSG example teaches us that not everyone has an understanding of
"legitimate change" that preserve data integrity, thus the EPSG effort
to clarify which changes are permitted.


> Another approach might be to use a 2-clause BSD license that, while
> permitting modification, does not allow the modified work to be
> distributed under the original name. But I would suggest that this is
> not necessary and just causes needless work from people who are making
> changes like fixing == to = to make shell scripts follow POSIX, who
> then have to think about whether their change is one that triggers the
> clause prohibiting using the name for modified works.
>
I think that protecting the original name could help. But I'm not sure
to understand correctly if in above paragraph, "not necessary" means
"because we do not expect peoples to do that" or "because the
consequences are not considered important"?


> The real way to ensure integrity of geodetic calculation is to publish
> test vectors and perhaps code, and encourage those to be added to unit
> tests in software that uses the transforms. Again these can be CC0 --
> no one is going to pay attention to modified test vectors.
>
Right - this is also what EPSG tried to do with Geospatial Integrity of
Geoscience Software (GIGS) tests. I would support moving forward with
this approach and I can help, but this could be the subject of another
thread.

    Regards,

        Martin