Message bases over 2gb

support multiple data files for >2GB total data. one file per message? Example:

        data/mail.shd
        data/mail.sid
        data/mail.sha
        data/mail/00000000.msg
        data/mail/00000001.msg

(advantage: no allocation issues... no .sda file) (disadvantage: waste clusters, large number of directory entries) Must back-up entire sub-directory of individual message files (for mail)? Can't use message number for filename, shared data among headers. Use random number to reduce fexist() looping. Maybe high-bit in offset indicates separate file? Only use separate file for > x-blocks (defined in message base header somewhere?)

multiple large data files? Example:

        data/mail.shd
        data/mail.sha
        data/mail.sid
        data/mail_0.sdt
        data/mail_0.sda
        data/mail_1.sdt
        data/mail_1.sda

(Advantages: retain existing logic as much as possible) (Disadvantages: back-up filenames: mail_0.0.sdt)?

(from todolist)