I can confirm that all the examples I gave are of ‘real’ Dutch persons (and
not the rare ambiguous kind like Erik). And clearly the American journals I
linked to want these articles prepended.
This is what’s not clear to me though; that U.S. journals explicitly
expect this. Where does you clarity on this come from? Is it just a
commonly understood expectation?
Presumably Rintze is speaking from immediate personal experience. One
of his example is his own article, published in an American scientific
journal, and has the references of Dutch authors sorted first from the
preposition part, then from the family name part:
http://aem.asm.org/cgi/content/full/74/9/2766#REFERENCES
Right, but all sorts of things happen in the publication chain that
are not the result of any clear set of rules. E.g. any one publication
is not any particular evidence of what the journal guidelines specify.
But that aside, this isn’t the most critical issue; I don’t have a
strong opinion on including this parameter, or not. But I do want us
to clarify the precise issues here.
There seem to be four possible cases for a style rule:
(1) always use particle in sort
(2) never use particle in sort
(3) want to use particle in sort, defer to personal pref
(4) do not want to use particle in sort, defer to personal pref
Before we move on, what do you mean by “personal pref”? Which “person”?
It looks like case (4) is illusory. An author who strongly prefers
that his name be sorted on the particle will just treat it as part of
his family name (as Bruce indicates). So that leaves cases (1), (2)
and (3) to deal with.
The Dutch publications he cites are of type (2) (sort on whatever each
person considers to be their family name).
Rintze’s own publication seems to be either of type (1) or of type
(3). Whichever it is, though, a (non-localized) style option is
clearly needed.
Since there are three potential cases, the safest course might be to
provide three values for use with a name-sort-with-infix (I still like
infix! 
option: “always”, “never”, and “prefer”.
Again, we need to be precise here: what “use” are you speaking of
here? Put differently, are you talking about CSL, or the data input?
That would leave an unresolved issue over how to signal the preference
of the named person in the data. The preference in favor of sorting
on the particle is easy, it just gets merged with the family name.
The preference against sorting on the particle is harder. Do we
need to make a “no-infix-sort” flag available in the data bundle for
each name?
Here you seem to be talking about the data?
I don’t really think this is a problem. If we have “Alexander von
Humbolt,” then I’d call the “von” an “article” (or just leave it out
entirely frankly), which by default means it is not included in the
sort. If the style expects to sort on articles, I’d set that in the
style.
So I see only two cases here, I guess. But it’s the end of my day, and
I’ve been focused on other things, so feel free to set me straight.
Bruce