Some way to carry metadata inside or along with Word documents

[simon]

See also #921

From #945, which I accidentally duped:

When you open a Word doc with Endnote references you can see XML or XML like tags that allow the references to be extracted from the document. It would be great if whatever changes we work into the Word plugin could accommodate this kind of extraction.

Biomed Central will give a discount to Zotero users who submit articles to them if we can offer this same kind of feature.

[simon]

Current thoughts are:

1. One solution involving shared collections, such that metadata is updated across users.
2. Another solution involving embedding the full metadata in the file along with user ID and key. When another user opens the library, s/he can import items or link references to items in his/her library. No automatic updating across users.

[stakats]

If we go with #2, how would it work? For example, maybe we might need to associate multiple user URIs with citations (e.g. ​http://zotero.org/users/stakats/AWMDSLR7). If I shared a doc with another person, she could either add that item to her library or link that citation to an existing item in her library. The citation would then refresh according to the metadata in her library. If she returned the doc to me, the citation would refresh to my associated item's metadata.

I'm not entirely sure what the objection to #1 would be, since both #1 and #2 require users to have Zotero. And for URIs to work properly, they both might need server accounts, too.

[simon]

The obvious objection to 1 is that I might have metadata I don't want to put on the Zotero server. While we can hope to assuage people's privacy concerns, I think this will always be an issue.

For 2 to work, we would definitely need to associate multiple user URIs. Users wouldn't necessarily need server accounts, since we can probably generate a unique ID without interacting with the server. Zotero isn't strictly necessary either, since theoretically any bibliographic management software that could read our format could use its own URIs, but I'm not sure this is really something we need to be concerned about ATM.

[dstillman]

Here are some assumptions we can probably make:

1) Local data is always authoritative. Changing an item in your library and refreshing the document changes the linked citation, even if the citation originally came from another user.

2) We more or less leave the problem of syncing data to users if they're not using server sync support or shared collections.

3) Exact matches, not counting userID/itemID/key/dateAdded/dateModified, with one and only one item in the library get automatically linked without user interaction.

4) We might help the user link citations to items with different metadata but the same external ids (URL, DOI, ISBN, ISSN, etc.), but it wouldn't be automatic. The same process would apply to items without external ids (where duplicate detection logic could be used) or items with matching metadata but that existed more than once in the library. It's also closely tied to #921.

5) It would be nice to offer some way for a user to open a document from another user and have the option of overwriting local linked items with the metadata from the document if any of the metadata was different.