REM%MIT-MC@sri-unix.UUCP (05/26/83)
Relay-Version:version B 2.10 5/3/83; site harpo.UUCP Message-ID:<1566@sri-arpa.UUCP> Date:Wed, 25-May-83 22:27:00 EDT From: Robert Elton Maas <REM @ MIT-MC> Regarding the 50-second pause from asking for an appliation program before it is loaded and can be used... It seems to me the problem is the whole application is loaded en masse, even parts that aren't needed immediately or maybe aren't needed at all in an entire use of the program. Perhaps the Lisa system should be modified to permit incremental loading, i.e. load-on-demand as in LISP's AUTOLOAD or MainSail's module swapping. Then the bare toplevel needed immediately could be loaded immediately, the user could select an option, a little more could be loaded, etc. so the 50-second delay would be distributed over several user interactions, and if some parts aren't loaded at all the total would be less. Perhaps pre-fetching could be done when idle, so that if the user waits 10 seconds after loading is done before issuing the first command 10 seconds of pre-fetching could already be done, and if the pre-fetch happened to match what the user needed there'd be no delay at all at that point. If command and program structure is treelike, with each module connected to a few inferiors and one or perhaps a few parents, the pre-fetch could simple fetch all immediate inferiors to the currently-displayed module. Anybody out there in Apple working on this kind of idea?