On Jul 23, 1994 Paul Edwards wrote in a message to Malcolm Miles:

 PE>> What percentage of applications rely on undocumented dos
 PE>> calls?  I'm not asking for you to say "Sidekick", I'm after
 PE>> a number.

PE> Well, I'd be happy with a guess.  10%?

Can't even guess. Start with most TSRs. Andrew Schulman lists some apps in
his book: the DOS commands SUBST, JOIN, PRINT and CHKDSK, also Windows,
SideKick, DesqView, Manifest and Norton Utilities.

Unfortunately no-one is going to say that their app uses un-documented
calls even if you could define what one is. What is un-documented anyway? A
lot of the functions not documented in the official MS-DOS reference are
documented in other Microsoft publications such as Microsoft Systems
Journal and the MS-DOS Encyclopaedia. Also much of the un-documented stuff
is pretty well known anyway so most people would not consider it
un-documented. On the other hand some of the information that is documented
is not correct or not complete. I remember back in the old Concurrent DOS
days you just had to try each app and see whether it worked.

PE> BTW, given this reduced scope of problem, how many hours of
PE> work do you think it would take to achieve this?  Just
PE> IO.SYS, MSDOS.SYS and COMMAND.COM remember.

Not being a system programmer I really couldn't guess the hours this would
take. Depends on how much of this sort of system level stuff you have done
before. I recommend that you get hold of Andrew Schulman's book
"Undocumented DOS. It is a very interesting read. It even includes a
utility, INTRSPY, that among other things can log information to a file
every time a program makes an undocumented call.

mgm

--- timEd 1.00+