Hi all, Martin Mueller asked on the list for recommendations on good uses of xml:ids, which resulted in a good discussion and a FR. I summarized the main suggestions on the ticket on SourceForge, but I post them here as well because I think we need to discuss this further, see below. The Guidelines already have some recommendations (numbering based on doc structure): * The Guidelines already suggest numbering based on doc structure ( http://www.tei-c.org/release/doc/tei-p5-doc/en/html/CO.html#CORS2) * Use 3 letters (e.g. from the title)+ 3 digits, incremental. E.g. HOL001, HOL002, etc.. * Same as above, but no fix number of digits. E.g. HOL1, HOL2, etc. * Prefix an id with name of element (this is simpler version of what the Guidelines already recommend) * Give tei:TEI an id and prefix every other id in the document with it (to guarantee cross-corpus uniqueness) These are all reasonable suggestions and there can be plenty more - which is why I think that the TEI should *not* give any recommendation on best practices for xml:id because it's a project management issue, not an encoding one. Myself, I prefer random ids, a practice that avoids introducing yet another level of complexity and data management. I understand the human readability, but sequences are too easily broken. And when parsing, relying on ID content instead of TEI content sounds like a bad idea.