Which attribute should be used on <biblioentry> to identify the type of resource? Possibilities include, but are not limited to: pubwork, role, typeof. I suggest allowing pubwork and expanding the current enumeration list.
Should there be a requirement that contributors (authors, editors, etc.) be grouped in the <authorgroup> element? Processors can handle either case, so it may be fine to allow both.
Should <abbrev> be required, and if so, what conventions should be applied to its contents? Currently, the default stylesheets use <abbrev> to generate inline references and, in some cases, as a label in the bibliography. They use xml:id as a fallback when <abbrev> is missing. There is enough variety in what different styles require for an inline reference that most implementations will probably need to generate inline references using the author and date information. The question is whether <abbrev>, if present, should override any generated inline reference or not.
Do the <citebiblioid> and <citerefentry> elements belong as direct children of <biblioentry>, and if so, how would they be used?
Should we use <othername> for middle name and initials, or leave them as part of <firstname>? The latter seems most expedient. However, <othername> could be useful for nicknames.
When a resource is part of another resource, should there be two <biblioset> elements as children of <biblioentry>, one for the direct resource and one for the containing resource, or should there be just a <biblioset> for the containing resource? My suggestion is the latter so that the main <biblioentry> and its direct children pertain to the resource in question and the containing resource is handled in <biblioset>.
Is it necessary to restrict <pubdate> to a particular format, such as xs:date? That may be necessary if the objective is to allow the stylesheets to use full or abbreviated month names or to select how much of a date to use. Since a publication date needs no more detail than YYYY-MM-DD, allowing that plus YYYY-MM and YYYY ought to be enough and be easy enough for writers to use.
Should the child elements of <address> be required? I’m inclined to recommend them, but not require that they be used, with the understanding that some styles may require the address in a form that doesn’t match a raw text rendering. Most of the time, however, I suspect this won’t be an issue, since most publisher addresses are just city, city+state, or city+country.
What, if any, requirements should there be on the <edition> element?
Simplest would be to just have a number.
However, there cases where the edition can be more than that. For example, “50th anniversary edition” or “Revised and updated edition.”
I suggest that if the edition is a number, the style can choose how to render it, and if the edition is text other than a number, the style should use it as is.
What, if any, requirements should there be on the <volumenum> element for a book?
Simplest would be to just have a number.
In periodicals, that should work fine.
However, for books, there is the case where <volumenum> could mean the number of volumes (e.g., 3 vols.) or have a more complex structure (e.g., Volume 4 of 5).
I suggest that when the value is just a number, it be interpreted by default as the volume number, but if the role attribute has a value of “numberof”, interpret it as the number of volumes.
If the value is text other than a number, the style should use it as is.
If <volumenum> is inside a <biblioset> element, it should only be a number representing the volume number of the resource being described.