Re: [boost] [GGL] Bost.Polygon (GTL) vs GGL - rationale

Subject: Re: [boost] [GGL] Bost.Polygon (GTL) vs GGL - rationale
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2009-11-16 07:16:13

> -----Original Message-----
> From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]] On
Behalf Of
> Jose
> Sent: Sunday, November 15, 2009 9:54 AM
> To: boost_at_[hidden]
> Subject: [boost] [Boost] [GGL] Bost.Polygon (GTL) vs GGL - rationale
> 
> I am trying to understand the goemetry libraries.
> 
> I find the currently under review GGL very useful for non-cartesian
> geometries (and maybe cartesian too!) but I don't understand what was
> the rationale for accepting GTL, which is a library mainly used in
> chip design. The reviews for GTL clearly stated that the library was
> not appropiate for GIS-like applications as its focus is mainly 2D.
> 
> So, can anybody explain how these two libraries fit together. I see in
> which cases GGL will be useful but I am rather confused with
> Boost.Polygon, as it's optimized for VLSI design (as stated in the
> BoostCon paper and by some of the reviewers)
> 
> Did someone decide that it was not possible to have a single GENERIC
> GEOMETRY library and we should have two ?

I think this about to be decided (implicitly - if GGL is accepted too, as
reviews so far might suggest).

FWIW, I concluded from the long discussions that although it would be neat to
have one generic library, it wasn't quite time for that because there were too
many unanswered difficulties (mainly about robustness and floating-point) in
using GGL to implement GTL.  And using a generic system might always require too
big a compromise on performance, as the regrettably acrimonious dispute over
benchmarks shows. (I wonder if the acrimony has been inflamed by the fear that
only one could get the 'prize' of being accepted into Boost.)

The problem domains appear to me to have quite different needs for speed,
accuracy and robustness.

FWIW, I see no reason why the two should not coexist in Boost, until such time
as it is found possible to implement GTL efficiently enough using a generic
library.  

Let users vote with their mice ;-)

Paul

---
Paul A. Bristow
Prizet Farmhouse
Kendal, UK   LA8 8AB
[phone], mobile [phone]
pbristow_at_[hidden]