[postgis-devel] Making SFCGAL mandatory

Regina Obe lr at pcorp.us
Tue Nov 13 20:21:26 PST 2018


> > I thought it was already defaulted to off?
> 
> configure.ac sets it to auto.
> 
> > Though I guess if we released 2.5 as default on "if found" we can't
> > change that.
> 
> It's a bug fix :-)  Anybody that wants it just has to add --enable, and
many
> others will be spared an accidental dependency.
[Regina Obe] 
Hmm I guess I didn't notice it cause I always build --with-sfcgal

I haven't released 2.5.1 yet.  Got caught off guard this weekend with other
emergencies.
Would anyone mind if I turned it back off before I push out 2.5.1
Which probably won't happen until end this week at the rate my week is
going.

> 
> > Correct.  It very well might happen before we release 3.0, but my
> > feeling is it's too early to make that call.
> 
> Hard to believe.  Being stable requires a multi-year track record.
[Regina Obe] 

?  not sure what you mean by this - doesn't every stable thing require a
multi-year track record?
After all PostGIS wasn't really considered stable 
until 1.0 which as I recall took 2-3 years from inception and even then it
was quasi-stable.

Stability is such a relative thing.  In a way for many use-cases SFCGAL is
plenty stable.  It's stable enough for me cause I don't include the guis in
packaging.
However it's less stable to me than the rest of PostGIS code base, because
we have less control over it.
SFCGAL we have little control over it as no active PostGIS maintainer works
directly on the SFCGAL code base and all of the functions are direct stubs
against SFCGAL
Which is also at the mercy of Boost and CGAL which I know nothing of aside
from my earshot hearing of people screaming about CGAL every-time they do a
release and break stuff.

Raster is part of our code base and Even Rouault is very responsive to GDAL
issues if we have any which we rarely do (at least ones that are easy to
debug or don't depend on some deep 3rd party plugin).

GEOS in theory is kind of like SFCGAL without all the extra dependencies one
can't control and the fact that  most GEOS maintainers are part of PostGIS
group 
(so it's easy to yell over there at someone or lift a finger myself when
people don't act fast enough).

Thanks,
Regina



 



More information about the postgis-devel mailing list