[Dune] Docu

[Robert Kloefkorn]

Hi Jö,

nice description. Could you please add this to the documentation 
somehow. I think we had a discussion about this behavior more then once 
now.

Best,

R

===========================================================

When computing the expected numbers, are you aware that

  - ghost entities have a limited interface (i.e. I am not certain 
whether all
    codim > 0 entities are present for a given codim = 0 ghost entity),

  - in ALUGrid some of the ghost entities are duplicates of each other, 
e.g. in
    the following cases:

    g: ghost codim=dim entity
    b: border codim=dim entity
    i: interior codim=dim entity
    G: ghost codim=0 entity
    I: interior codim=0 entity

     +-----+-----+
     |     |     |
     |  G  |  I  |
     |     |     |
     +-----+-----+
     |     |     |
     |  I  |  I  |
     |     |     |
     +-----+-----+

    The ghost entity actually appears twice, once for each interior 
entity it
    is a neighbor of.


     g-----g-----g
     |     |     |
     |  G  |  G  |
     |     |     |
     b-----b-----b
     |     |     |
     |  I  |  I  |
     |     |     |
     i-----i-----i

    AFAIK the middle ghost vertex appears twice.


If that doesn't help, can we get some numbers here?  Actual vs. expected
element counts ?  Serial vs. parallel?