[geos-devel] deprecate non-thread-safe CAPI interfaces

[Roger Bivand]

On Wed, 10 Feb 2010, Howard Butler wrote:

>
> On Feb 10, 2010, at 5:26 PM, strk wrote:
>
>> I'd like to deprecate non-thread-safe version
>> of CAPI interface in next release, and stop providing
>> them for new methods.
>>
>> If anyone sees a reason not to do that please scream
>> before it's too late (you've plenty of time).
>
> The proposal would be to stop providing new methods and method updates 
> for the non-thread-safe API, correct?  While I'm generally in support of 
> this, I think it is one release too early.  I think that people are just 
> now undertaking the development efforts to use the thread safe API in 
> their software, and we should give folks a bit more time before we pull 
> the rug out.  Let's wait until 3.4 instead of doing it in 3.3.

My understanding is that MSYS builds for Win 32 (and maybe 64) do not have 
or use pthreads. R is not thread safe and will never be (compute threads 
are handled differently by say threaded BLAS or coarse-grained partition 
of tasks into sections for parallel computation). I think that GDAL 
provides a MUTEX_NONE stub which I see kicking in on MSYS.

On Linux and OSX, this probably isn't an issue, but building static GEOS 
for Windows necessarily using the R build train (MinGW, MSYS) is fragile, 
and could be broken by hasty changes. The potential R install base for 
GEOS functionality is at least in the thousands, mostly Windows users 
(sorry), who rely on the developers providing packaged binaries through 
CRAN.

If you could push this a little forward in time, and let me try with the 
development version to make sure that any transition can be managed, I'd 
be grateful. As I've said before, the motivation for rgeos is to provide 
an alternative to non-free GPC that has been interfaced to R.

I realise that GEOS usually goes in at the base of the software stack, and 
usually interfaces GIS-type systems, so this case is somewhat atypical.

Roger

>
> Howard
>
>
> _______________________________________________
> geos-devel mailing list
> geos-devel at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/geos-devel
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no