This response springs from an issue raised during the discussion of "active" dates but it actually deals only with firm dates.

Interesting question, Richard.  Some will know that in order to test a module that may be used to generate 046 fields from information in the 670 fields, the authority loader we use here has begun to compare the 046 present in incoming LC/NACO authority records with information pulled by the new module from the 670s, and to report the discrepancies.  (Yes, there are plenty of these.)  As it happens, this module refuses to proceed if the authority record contains 046 and 100 $d and the years in the two disagree, but until now I haven't actually used this for anything.

I just had a test program run through the most recently issued names file (14.17), pulling out records with 046 fields and comparing them to the 100 $d.  The program found 2,945 incoming non-delete records with 046 fields containing either $f or $g (or both).  Of these, the program was able to handle 2,727 successfully. Of the 24 records rejected for one cause or another, 16 were rejected because information in the 046 field disagrees with information in 100 $d. Make of that proportion what you will.  Here are more examples than anyone will need (I've deliberately eliminated the text from 100 $a so we can focus on the numbers and not worry about side-issues):

046:  : |f 1900 |g 1969

100:1 : |d 1900-1968

670 field: died 1968

046:  : |f 18200929 |g 19080113

100:1 : |d 1820-1909

               670 fields show both 1908 and 1909

046:  : |f 18991018 |g 19781029

100:1 : |d 1889-1978

                670 fields have 1889 everywhere

046:  : |f 1922 |g 1965

100:1 : |d 1921-1965

                670 fields have 1922 everywhere

046:  : |f 1987

100:1 : |d 1986-

                670 field says born 1987

046:  : |f 1946

100:1 : |d 1904-

                670 field says 1904

046:  : |f 19480708

100:1 : |d 1965-

                670 field says July 8, 1965

046:  : |f 10600927

100:1 : |d 1960-

                670 field says Sept. 27, 1960

046:  : |f 19160811 |g 20110522

100:1 : |d 1916-1985

                670 fields only say 1985

046:  : |f 18590109 |g 18380817

100:1 : |d 1859-1938

"For yourself, sir, should be as old as I am, if like a crab you could go backward."

(Sorry; couldn't resist.)

As far as I can tell, most of the above stem from one kind of operator error or another, either simple typographical errors, or perhaps the eye skipping to the wrong part of the record or even another record.  Errors of this type will continue to happen regardless of legislation.  (But it does seem clear that this test needs to be added to the authority loader as well, so that typographical errors can be corrected.)

At least one of the records, though, involves a disagreement about a date in sources, and having a clear statement directing people what to put into the 046 would be a very good thing.  Here's another record with the same condition; in this case, the person constructing the record attempted to describe the problem, instead of resolving it (in the original record, each of these 046 fields also has subfield $v):

046:  : |f 1873 |g 1968

046:  : |f 1874

046:  : |f 18740330 |g 19680911

100:1 : |d 1873-1968

Against this, maybe it would be confusing in some way if 046 and 100 $d did not match. Maybe it’s something on which PSD could give an opinion?