Hi Benny,
I would say the bug is with Rootsmagic, as the spec clearly says:

 NAME_PIECE_SURNAME: = {Size=1:120}
Surname or family name. Different surnames are separated by a comma.

From: http://homepages.rootsweb.ancestry.com/~pmcbride/gedcom/55gcch2.htm#NAME_PIECE_SURNAME
H'm, yes, that's a part I didn't read. Pretty convincing, not?

Well, read this:

Surname prefix or article used in a family name. Different surname articles are separated by a comma, for example in the name "de la Cruz", this value would be "de, la".

I found it right below the paragraph you quoted above. It says that de and la belong together. Nice! How would you distribute those when you have two surnames, and your theory is right? :-)


Should be used to know how the pieces are concatenated, and to know what pieces there are the subfields (2).
If RootsMagic does not do it like that, they don't follow the Gedcom 5.5 (in my interpretation, I probably wrote that code somewhere in the past...)
See also http://homepages.rootsweb.ancestry.com/~pmcbride/gedcom/55gcch2.htm#NAME_PERSONAL which indicates this part is the correct form, with the other fields usable to parse this field
So, you say, programs must follow the NAME_PERSONAL field, because that's the correct form, right? I agree.

But if that's right, and I think it is, we have a new problem:

When I import the gedcom created by Gramps into Gramps, I see that the parts are imported in sequence, as if they form a single name, and the comma's are imported too. So, on the next export, Fatima looks like this:

1 NAME Fatima /, de Moreiro, Melo/
2 GIVN Fatima
2 SPFX , de
2 SURN Moreiro, Melo

That's not what I read here: http://en.wikipedia.org/wiki/Fatima_Moreira_de_Melo

Conclusion: Gramps does not follow GEDCOM 5.5 either, so I definitely need to make that bug report.



P.S. I hope you forgive me for misreading Moreira as Moreiro.