I am collecting books and stuff for a similar to project alice,, (massive data upload of literature seems interesting to me for a ai if based on vml, soo my collection is large but I also spidered webpages and pdf'ed them too, and well I just like have a private archive to search and sift as we get connection problems up here and websplits often. I sometimes play the ebooks as audio, (depending on the format which tools I use if at home I use a winxp machine with msreader read out loud via .lit files or, pdf read out loud (adobe acrobat though been considering trying some other stuff) my linux machine mainly a fedora 14 uses carnival/ festival, (I also must confess the linux machine is running my calibre, I initially had trouble around 6500 books or items when using winxp (even sp3) but with linux, WOW! my speed and stability of calibre is vastly improved, when importing -add more then 10000 items at a time, I do get a pile of errors from drm to interpretor stuff, but I have not crashed it yet, (with linux)
If I am adding a huge collection I break it to smaller chunks 10 gig seems nice, then run flint across to detect dupes then fdupes maybe depending on my patience,,
I like the idea of converting all to epub, then if something happens I can just copy or move out all epubs if I have to rebuild (plus this cleans out any drme'd mistakes which inadvertently got added (some tools for downloading or syncing do funny things from signatures to read write permissions,,, so,,hahah)
I really like calibre extremely ambitious, all in one swiss army knife for documents,, and last year or so stability has really spun up nicely,,, (not sure if I would try a windows server again but maybe if I was gonna suggest anything, laguage translation (tough as heck and nearly impossible but this would really round out my thoughts, (and perhaps a option to delete the original files on a import after successful copy? (just the option would be nice ) but great program actually a server in its own right, I would love to know how others test this,, ( I am sure a single file would be quick for testing import and conversion, but yes even huge archives seem stable now to me :-) even if they seem pointless, and if a Aritificial Intelligence was ever developed, hmm some of these archives would be fun to upload or recode?