Hi All,
This has been a really great thread. Here some more thoughts; "Having your cake and eating is also possible with some ingenuity and elbow grease."
While JPEG2000 seems like a great idea on paper, it is often not used as the primary storage format due to concerns about lack of support in the browser (or on the desktop for that matter). A number of years ago I was working on an image portal solution at the TU Delft in the Netherlands, and made a decision to use JPEG2000 as the storage format, but convert the files to JPEG 'on-demand' linked to an image viewer. This is similar to what is done with large TIFF originals in many other projects, but the big benefit was the speed at which one could do arbitrary resolution segments from a JPEG2000 file.
For the viewer I chose the MS Seadragon Ajax version (it has lovely smooth scrolling and zooming, and cross-browser support) so no Flash was needed. In stead of pre-tiling the images (like Zoomify does, or the MS Seadragon back-end tools) I wrote a Python script that serves up the needed zoomed tiles on demand via HTTP, in effect 'pretending' to be a Seadragon back-end. It worked really well, and I was quite satisfied with the solution. For the JPEG2000 processing, we purchased a license from http://www.kakadusoftware.com/ and used the command-line tools called from Python.
Here is an example of this solution in action: http://repository.tudelft.nl/view/ir/uuid%3Aa5c3103d-e20d-4a3c-9eb6-2dc7b287...
(this is the institutional repository showing some sample pages from PDF research papers tiled for preview before downloading as some files can be hundreds of megabytes, but imagine it being a manuscript illustration of you will)
Hope this is of use as an example to others.
Etienne Posthumus Amsterdam, The Netherlands