Hello!

Regarding
http://www.loc.gov/marc/marbi/2012/2012-03.html I noticed that the
subfield $i is aimed to contain an "m" when the record is

> fully machine-generated

I consider that this could be a little more granular and we could offer
the possibility to store the relationship (at the time of the assignment)
between the original classification and the machine-generated one. We
could use "eq" for "equal" and likewise "nt", "bt", "rt" (or similar). One
of the examples in the 2012-03 proposal is
http://export.libris.kb.se/DS/ which provides similar information with
symbols.

We can not be sure that the classification translator software will be
available in the future and that such information can be found there. If
it is available, it may have been improved (or otherwise modified) and
could provide another classification suggestion. Therefore, it could be
useful to store the relationship between the original and the
machine-generated classification (at the time of the assignment). In the
case of Libris (as examplified in MARBI 2012-03), this relationship is
provided when the site is used. Of course, we can not have any garantee
that the relationship provided by the software is nor will remain accurate
...

Nonetheless, I consider that such information could be useful for the
retrieval algorithm.

Regards!

Saašha,