The ideal solution would be a tiny bit of automation (perhaps created by a student if techie stuff isn't your thing).
"tiny" is a very misleading term, i think. unless you can show me this "tiny" thing.
All that's required is a tab-delimitted file (or database or spreadsheet) with 2 columns: id, annotation. Now, in your favorite scripting or programming language, iterate thru the file, read the id and annotation, then replace the former with the latter in the marked-up book. (Including the appropriate (X)HTML wrapper, as noted.) For someone who doesn't write scripts, it's not trivial. For someone who does, it's a few lines of code. Note that with a little more techie work, the process could be simplified for the annotaters. They could add the annotation text directly in the document, surrounded by unique delimiters. Then, a script could generate any version, e.g. replace delimiters with (X)HTML wrapper and/or with a generated unique ID; extract the annotations to a separate (X)HTML file that can be printed on its own, etc. All this stuff is pretty easy for a college student with any scripting experience.
Then you could keep the annotations separate, and just add small markers to the original text.
um, keeping the annotations separate is a good idea. but requiring "small markers" in the original text is not. the text should remain unchanged, for many reasons.
Sure, in the ideal world. Meanwhile, inserting unique IDs is a pragmatic solution. Of course there's some work involved to sync the marked version with any future PG updates, but given "diff" tools, it's not that much work for a handful of annotated books. (And, with a little work, could be largely automated.) And, I'm all in favor of someone taking the time to get an XPATH solution working. -- Cheers, Scott S. Lawton http://Classicosm.com/ - classic books http://ProductArchitect.com/ - consulting