New API for models of MeshDomain_3 based on 3D images

My plan is to deprecate CGAL::Gray_image_mesh_domain_3 and CGAL::Labeled_image_mesh_domain_3 in favor of new constructors of CGAL::Labeled_mesh_domain_3.

The API of those constructors would be similar (I simplify) to:

template <class GT>
class Labeled_image_mesh_domain_3 {
public:
  template <typename Functor>
  Labeled_image_mesh_domain_3(CGAL::Image_3 image,
                              Functor image_values_to_subdomain_indices,
                              double relative_error_bound = 1e-6);
  // [...]
};

The problem is that I need to distinguish between:

• a gray image for which I will use trilinear interpolation, and
• a labeled image, for which I will use no interpolation, or a labeled trilinear interpolation.

So the new constructors will use a tag:

template <class GT>
class Labeled_image_mesh_domain_3 {
public:
  template <typename Functor>
  Labeled_image_mesh_domain_3(CGAL::Image_3 image,
                              CGAL::Gray_image_mesh_domain_tag,
                              Functor image_values_to_subdomain_indices,
                              double relative_error_bound = 1e-6);

  template <typename Functor>
  Labeled_image_mesh_domain_3(CGAL::Image_3 image,
                              CGAL::Labeled_image_mesh_domain_tag,
                              Functor image_values_to_subdomain_indices,
                              double relative_error_bound = 1e-6);
  // [...]
};

The precise name of those tags does not matter.

Alternatively, I can use named constructors, with static class functions acting as constructors:

template <class GT>
class Labeled_image_mesh_domain_3 {
public:
  template <typename Functor>
  static
  Labeled_image_mesh_domain_3<GT>
  create_gray_image_mesh_domain(CGAL::Image_3 image,
                                Functor image_values_to_subdomain_indices,
                                double relative_error_bound = 1e-6);

  template <typename Functor>
  static
  Labeled_image_mesh_domain_3<GT>
  create_labeled_image_mesh_domain(CGAL::Image_3 image,
                                   Functor image_values_to_subdomain_indices,
                                   double relative_error_bound = 1e-6);
  // [...]
};

Which solution do you prefer:

1. constructors using a tag, or
2. static class functions acting as named constructors?

I personally prefer the named constructors:

• their names map easily from the named of the deprecated classes they replaced,
• they are easier to implement (because there are a lot more parameters, and I use the Boost Parameters library),

but as far as I know that would be the first time we use the named constructors pattern.

Concrete usage would be:

typedef CGAL::Labeled_mesh_domain_3<CGAL::Epick> Domain;

// using named constructors:
Domain domain = Domain::create_gray_image_mesh_domain(my_image, my_functor);

// using the constructors with tags:
Domain domain(my_image, CGAL::Gray_image_mesh_domain_tag(), my_functor);