[Benchmarking] library benchmarks in benchmarks

[Mateusz Loskot]

Frank Warmerdam wrote:
>>>> We have developed benchmarks to test performance (and functionality) of
>>>> (Open Source) GIS / Geometry libraries.
>>>>
>>>> Originally the tests were created to compare our library (GGL, recently
>>>> accepted by Boost and renamed to Boost.Geometry) to other libraries,
>>>> with respect to performance and functionality.
>>>>
>>>> However, the test suite is broader than that because it can also be
>>>> used
>>>> to compare for example GEOS and CGAL. Therefore we propose to move
>>>> it to
>>>> the SVN http://svn.osgeo.org/osgeo/foss4g/benchmarking
>>>> <http://svn.osgeo.org/osgeo/foss4g/benchmarking/>, it seems to fit
>>>> there
>>>> perfectly.
>>>>
>>>> It currently compares C++ libraries (geos, cgal, gpc, wykobi, gtl,
>>>> terralib, ggl) and a Java library (jts), for the functions area,
>>>> centroid, convex hull, intersection, clip (=intersection with
>>>> rectangle), simplify, within. But this can be extended.
>>>>
>>>> There is also a specific test, "starcomb", for more 'challenging'
>>>> intersections (done with a selection of the mentioned libraries only),
>>>> and "triangle" (intersecting polygons with 1000ths of triangular holes)
>>>>
>>>> I like to hear opinions if this is found a good idea, and if there are
>>>> possibly people interested in extending tests, performing tests and/or
>>>> maintenance.
> 
> Barend,
> 
> I don't have any objection to this being hosted in it's own subdirectory
> of the foss4g/benchmarking area as long as it isn't terribly large.  But
> it seems quite a distinct activity from the existing "web mapping"
> shootout.  I'm not convinced it would end up in the same presentation.

Frank,

Perhaps it's a little misunderstanding.
We do not aim to include the geometry performance results in any
presentation. We've just asked for hosting of this package.
The point is that the geometry performance package is supposed to be
project independent, as objective as possible, ideally if it other
developers could participate to improved it, and so on.
IMHO, OSGeo Community is best candidate to host this initiative,
unless there is a clear objection. Is there?

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net
Charter Member of OSGeo, http://osgeo.org