This is an old thread. But I do have an update.
I started off displaying uncompressed epub3 on a website with PHP. I got it to work for output for Calibre (and Jutoh for that matter). But my codes didn't work for all epub3.
Epub3 Best Practices uncompressed didn't work because its XML made relative paths between *.xhtml pages and images differently than calibre. I didn't like being tied to a GUI to make new content too. Sometimes I like to ssh into a site and then just use a text editor.
So I made a plugin (to my system) that shows HTML pages as book pages.
Top level directories are chapters. Next Page and Prev Page buttons set a "last page read" cookie you you can come back a week later and jump to your last read page.
A collapsable table of contents provides random access to all parts of the book without resetting the "last read page" cookie. So you can jump around and look at diagrams, and then return to your last read page. Right now or days later.
An searchable index is still missing. But it's in the works. I'll use Apache Lucene. I've worked with Lucene on a Java Servlet Engine. So I kinda sorta know how it works.
This code is not ready for distribution yet. Maybe soon. I'm working on it. It will be on Github.
https://montana-riverboats.com/?robopage=Birds