The thread asks "
How many people only (or mostly) use file mode due to library limitations?". My position is:---
(1) XBMC's library mode is so limited that people must either use file mode or be unhappy with XBMC,
but
(2) XBMC's file mode is itself so limited that people shall never grow too unhappy with the present state of library mode,
and
(3) at present it may be both easier and more desirable to improve upon file mode.
XBMC's library mode handles movies and tv shows nicely, but treats everything else (documentaries, on-line lectures, opera, ballet and theater performances, anime) as library orphans. These orphans may be treated well if misrepresented as movies or tv shows, but there are limits to how well this trick can work. And even movies and tv shows are not treated well enough: there is no way to categorize them more than one level deep. Sub-genres or sub-tags are out.
AFAIK the cause of these limitations lies in the fact that XBMC's database model is non-relational, or at least non-normalized relational (which is practically the same as non-relational). In a normalized relational database, it would be easy for anyone to edit the database structure to reflect his or her organizational needs. In XBMC, this structure is rigid. Now I confess to not being computer proficient enough to rule out the possibility that all is well with XBMC's database model and the fault lies with the screen drawing's or the interface's inabilities to do this or that, but I had a short exchange with Montellese some time ago in which he pointed out that the database is intentionally non-normalized because otherwise it would have been too slow, so I concluded that the fault lies there.
Sometimes using smart playlists can help overcome the database's limitations, sometimes it cannot, and sometimes a playlist can do even less than the database allows. I thought that a smart playlist is a way to express a standard SQL query in XML for the database to understand, but no, it doesn't function that way. It recently transpired that XBMC's (v12.x) database has implemented a function (video tags for tv shows) that the playlists have not. They would have to be told explicitly to understand tags, and how to do it, which has not happened in time for the release and has been postponed for v13.x. If they were simple SQL queries they would have been able to do whatever is allowed by the database.
Thus file mode (or playlists with path-based rules, which is of course file mode in disguise) is the default option of organizing videos in case the database mode fails you. It may even work well enough. But in a perfect mirroring of the library mode's limitations, file mode is limited too: you cannot look inside descriptive text inside the database to filter videos according to content. It is at present not possible to find all occurrences of the string "M-16 rifle" in the library, even though the correct nfo file, which contains the correct tag <plot>, which contains a correct description, has been correctly scanned into the library.
None of the above limits XBMC's functionality as a media center: we are talking about annoyances not defects. Of the two annoyances, I have come to the conclusion that the inability to do a full text search is the most disturbing. After all when I work with media files on my PC I neither have library mode nor get too excited about its omission. But the thought that my file manager could not select files based on text string content would drive me mad.
