@Jdforrester-WMF thoughts?

This particular tack seems like it would make the front end not very clean. For instance, the casual user probably would have no idea what "Harv" referred to.

Not sure if we could just put the harv citations in the re-use panel?

I think the idea is not to focus on Harv itself but to have a way to define a source (book, magazine) somewhere and insert that source by quoting the page.

At the moment, you can cite a book. If you want to cite a precise page for that book, you need to create a dedicated citation, and an other one for an other book. Have a way to define a Bibliography into the article, and then be able to create <ref> tags for each page would help.

But that is still a feature for more experienced users, I think too.

How should VisualEditor tell the difference between these two lists bibliographic citations on a page?

Publications

Alice Expert has written the following books:

• Expert, Alice (2012) ''The Sun is Really Hot''. Random House.
• Expert, Alice (2016) ''The Sun is Really Big''. HarperCollins.

Works cited

• Chris, Bob (2015) ''All about Alice Expert''. Simon & Schuster.
• Politician, Paul (2016) ''Why I Hired Alice Expert''. Hachette Livre.

When you're editing a BLP about an author or academic with a long list of publications, you really don't want it to give you a list of everything that's in a bulleted list or everything that uses a citation template.

• when VE detects a bulleted " {{Citation " it will be a citation for use with harv
• any bulleted citation can be toogled (while being created or edited) to be used or not for harv referencing, by adding or removing the " |ref=harv "
• when adding a new bulleted citation, " |ref=harv " (or not) gets inherited as a selection from just above, if just above it there is also a bulleted citation
• non-harv citations wil not appear in the list to choose from

I think that your first point is redundant. It sounds like you'd really be having it search for a particular parameter in a citation template.

A " {{Citation " citation should be treated as if it already contains a " |ref=harv " in code, because it behaves like it does even if the parameter is omitted. There are articles with 100, or more, " {{Citation " citations, used by harv referencing (example). I should be able to make a new harv ref easily, based on a citation -or a combination of citations- that already exist, like in the example given.

But if you assume that all bibliographic citations on the page are references that support article content, then pages that use the citation templates as part of content rather than references, e.g., https://en.wikipedia.org/wiki/Bibliography_of_World_War_I may be painful to edit.

This example does not contain any " {{Citation " citation. And should not. I cannot find an example why someone would like to make harv references out of such lists. I feel there is a misunderstanding here.

https://en.wikipedia.org/wiki/Template:Citation is similar to the Cite templates, but by default generates anchors for Harvard references whereas the Cite templates do not and one has to place the anchor manualy there, by adding " |ref=harv " . These " {{Citation " templates, if met inside an article, should be treated as if they had the " |ref=harv " parameter always set, meaning they should always appear in the selection list. Citations based on Cite templates could be toogled in and out from the list, by using the " |ref=harv " switch, while unformatted citations should not be used at all for harv referencing.

So you would use this only for {{Citation}} but not for the far more popular {{Cite book}} (which appears about 50 times on the page I linked in the previous comment).

So here are three facts:

• https://en.wikipedia.org/wiki/Gerardus_Mercator is one of the (exactly) 582 articles at the English Wikipedia that uses both {{citation}} and {{harv}}.
• https://en.wikipedia.org/wiki/Gerardus_Mercator#Books uses {{citation}} to display a list of books written by the subject.
• https://en.wikipedia.org/wiki/Gerardus_Mercator#Bibliography uses {{citation}} to display a list of the works that were cited as sources for the article.

Leaving aside the question of whether it makes sense to develop anything to support a citation system used in just 0.009% of the English Wikipedia's articles, what I want to know is whether you propose to magically have only the ==Bibliography== sources but not the ==Books== he wrote -- both of which are recorded in the wikitext with exactly the same wikitext template -- be available in this search list, or if you have a plan for displaying only the useful ones.

Only cite book containing ref=harv , cite journal containing ref=harv and citation bulleted citations should feed the list to select from to make harv references. The list of books a writer has written should not be cited by any of the three above, meaning that citation should not be used and cite book should be used without the ref=harv parameter/switch, there.

To easily keep lists clear from unneeded ref=harv, a toogle button for adding or removing ref=harv parameter should appear while editing cite book and cite journal bulleted citations. This toogle button should be disabled for harv citations already used for harv referencing.

The first item of a bulleted list of citations should make a memory for next items of the same list, for pre-selecting ref=harv or not while new list items (bulleted citations) are being created, to save clicks and mistakes. VE would this way recognize, by the first item, that this list should be treated for example as a list of books written by the writer and not as a harv citations list, and assist.

Easy Harv references might attract scientists in Wikipedia and would help articles having "owners" who revert whatever is not a harv citation. Programming all these might be costly for what we really need, but this is the analysis.

Theory:

The list of books a writer has written should not be cited by any of the three above,

Data:

===Books===
*{{citation
|last=Mercator |first=Gerardus |year= 1540
|title=Literarum latinarum, quas italicas,cursorias que vocant, scribendarum ratio (How to write the latin letters which they call italic or cursive)
|location= Antwerp
|oclc=63443530
|postscript = . Available online at the [http://www.rarebookroom.org/Control/mcrlit/index.html Library of Congress] and  [http://daten.digitale-sammlungen.de/~db/0002/bsb00026193/images/ Das Münchener Digitalisierungszentrum]. It may be [https://download.digitale-sammlungen.de/BOOKS/download.pl?id=00026193&nr=1 downloaded as a pdf] from the latter. This book is the subject of a monograph which includes a translation of the text {{harv|Osley|1969}}. 
}}

You are saying that this article "should" not use this citation template in the list of books a writer has written. I am saying this article "actually does" use this citation template in the list of books a writer has written. Unless you can prevent people from doing this, then it will not be possible to achieve that goal.

A good practice is to offer multiple lists to select harv citations from, (items grouped by ==section== in different lists) and prefer to automatically expand the list already containing most harv refs anchored on its items or the most recent one used during the edit. This will allow citing even the author's books, if needed.

T52769 does not seem to have anything to do with the current topic, at least on the surface. Is for both difficult to handle templates?

It seems to me that T151301 and specifically the task for VE at T151308: Make VisualEditor compatible with "extends" parameter would mitigate the need for this task.

I wonder if this is at least two very different things, the reference system in use for a ref mark and whether it is inlined.

A ref-mark might be formatted according to Harvard style. This could even be different from Harvard, and better yet, it could be a user preference. With a bit of Javascript it does not imply cache fragmentation.

A ref-mark can can also be inlined, that is to not put it into a superscript element. That could solve the quite awkward present situation, if we write about the referenced source, and not just use the reference to support a statement. Which reference system system is really not important in this case.