[postgis-devel] [geos-devel] GEOS C++ API deprecated? (was: PostGIS 2.5 what should be minimum requirements?)

[Mateusz Loskot]

On 1 October 2017 at 20:47, Sandro Santilli <strk at kbt.io> wrote:
> On Sat, Sep 30, 2017 at 09:26:42PM +0200, Mateusz Loskot wrote:
>> /topic changed
>> /cc geos-devel
>>
>> On 30 September 2017 at 20:47, Greg Troxel <gdt at lexort.com> wrote:
>
>> >   ### Using the C++ interface (discouraged)
>> >
>> >   NB: The C++ interface should not be used directly; the GEOS project
>> >   views it as a bug for another program to use the C++ interface or even
>> >   to directly link against the C++ library.
>> > [...]
>
>> > Mateusz Loskot <mateusz at loskot.net> writes:
>>
>> Moreover, this paragraph has no rights to be there or in any official
>> GEOS writing.
>> I'm very surprised Sandro allowed it in - I assume a merge in rush.
>
> Maybe "a bug" is too much, but the "discouraged" label is important.
> We don't want client software to use the C++ API,

So, we decide for clients?
Clients are warned, we don't care what API they decide to use.

> and you see the reason today (GEOS is kept back in Debian because a client used the
> C++ API).

So what? It's still ac client's authors freedom to decide.
If they use the C++ API despite the no stability promise policy,
and they packaged it for a distro, perhaps authors of the software
should not be doing what they are doing if they don't care about
reading the basic info about the library they use.

>> Finally, even if GEOS C++ API was/is marked as deprecated, then I ask
>> where is the RFC, where is the PSC voting the motion,
>> where is the public announcement?
>
> Idea was announced here:
> https://lists.osgeo.org/pipermail/geos-devel/2005-April/001375.html
>
> Work was announced here:
> https://lists.osgeo.org/pipermail/geos-devel/2005-September/001574.html
>
> First release and recommendation to avoid C++ API was here:
> https://lists.osgeo.org/pipermail/geos-devel/2005-November/001619.html

Those are not related to this particular discussion.
Those are about developing the C API, but its introduction had not
deprecated anything. It just made life of developers easier.

> I'm not aware of any motion to explicitly mark C++ API as "deprecated"
> but as it's effectively not maintained, it is continuosly "deprecated".

The C++ API is maintained, but every new release introduces new C++ API
and saying it is becoming deprecated is incorrect.

Unless, Sandro, your aim is to eventually mark C++ API deprecated
and stop installing C++ API libraries and headers.

That will require RFC and that is what I'm debating about.

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net