I've heard many conflicting reports about which input mode is preferable for e.g. the command description for in_mouse in the ezquake console - "2 - Direct Input (recommended for uberleet gaming performance (c))"

http://en.wikipedia.org/wiki/DirectInput - "It should be noted that Microsoft recommends that new applications make use of the Windows message loop for keyboard and mouse input instead of DirectInput"

Can someone explain precisely why one or the other is better?

Once I found article at MSDN where Microsoft explains different input methods but I can't find it now

If we talk about Windows and in_mouse then:

in_mouse 1 - worst
in_mouse 2 - better
in_mouse 3 - even more better, probably

Some links in my post in another thread - http://www.quakeworld.nu/forum/viewtopic.php?pid=37620#p37620

thanks, raw input feels good but only mouse1 and mouse2 are bindable and theres too many bugs in the beta ezquake : )

went back to direct input a week ago with ez 1.8.2, even with neg accel on fast movements, real battle is much easyer. bunny is easyer and lg aim went up like 8%. Its all becaouse off no delay on mousemovments. There is definitely some delay with in_mouse 1 and in_m_os_parameters 0. U get rid off it with in_mouse 1 and eiter 1 2 or 3 on in_m_os_parameters, but tracking doenst feel so good as 0, can even be difficult to just hold crosshair still.
Its also strange how jerky mouse is in windows when alt tabbed with qw running with in_mouse 1 and some os_parameters settings. is it this mousemovements qw reads, and it just look smother in qw with high fps smoothing it out?

Yesterday i installed ez 1.9 beta 2, and i just love in_di_bufsize, setting it to 128 or something almost remove all neg accel with in_mouse 2(just some on the fastest movments. ie when i try to just sweap it across mouse mat as fast as possible, witch may be due to my mouse max speed limit or something)
and one thing, qw is just more fun with direct input. Feels more connected

Whatever works for you is generally fine by me, but if your lg went up 8%, there is def something fucked up with your in_mouse 1 or your aim. Even though I consider dinput the worse option, I can honestly say that after some days of using it and getting used to it, the difference is not that big, and definitely nowhere near 8%, jeez.

yep placebo.

i rest my case, gotten around to test in_mouse 3, and its better than in_mouse 2 even with higer in_di_bufsize.
Its no delay like with in_mouse 1 and in parmeters 0. And there still some inconsistency with the tracking with in_mouse 2, the same movement will not always result in the same result. lg was easy with direct input but i it felt like i had to correct fast flip shots with rl, instead of just doing the flip.
In_mouse 3 is awsome