We spent a long time a year or so ago defining elements for citations and
ended up with an element <part> that was under <relatedItem>. Currently,
all MODS elements are defined under <relatedItem>, but in addition <part>
is ONLY defined under relatedItem.

The MODS guidelines say that <part> is limited to use for <relatedItem
type="host"> for generating citations about the location within a host or
parent item (although this can't be enforced by the schema).

There are some problems with this. One is that it breaks the content model
for relatedItem where it brings in all MODS elements, since <part> isn't a
top-level MODS element. The other is that, because MODS requires at least
one element, at least one MODS element must be under <relatedItem>, but
since <part> is not a MODS element you can't use <part> alone.

We have a large initiative called the National Digital Newspaper Program
to digitize newspaper pages from 1836-1923. We are planning the
architecture now and plan to use METS and MODS. There will be METS
documents for issues of newspapers with MODS metadata for the issue as
well as possibly for the pages (the pages will be detailed in the METS
structural map at least). There is a need to include information about the
enumeration (volume, issue, etc.) about the particular issue being
digitized which is what is being described. It seems more intuitive in
this instance to do this at the MODS level rather than the related item

So my proposal is to define <part> as a MODS element. The user could
choose whether to use it at the MODS level or related item level. It would
provide more flexibility and would allow relatedItem to be recursive and
include any MODS element. It would solve the problem of being required to
have one MODS element in addition to <part>. And it wouldn't invalidate
any existing records.