Dan,
Please try to appreciate the position that we are in.
Every large piece of software has to have defined interfaces and
undefined implementation details. (This is separate from the notion of
publicly visible source code. Even open source projects have details
that may change without affecting the defined interfaces.) If none of
the interfaces are defined, then you get unstructured chaos. If all of
the implementation is defined, then you can never change or extend it.
Every statement we make here can be taken by lawyers to mean that we are
implying that the OS is guaranteed to do one thing or another. So, if
we know that the details that you are asking for are in the undefined
range, which means that they may change from release to release, then we
are literally legally constrained from answering your question. If we
tell you that it works a certain way, and then you build software that
takes advantage of this knowledge, then it’s quite likely that you’ll be
frustrated when a new version is released (perhaps even just a service
pack) and your software doesn’t work.
If you want to know about how the OS works from an academic viewpoint,
then, by all means, dig in with a debugger and figure it out. Read the
“undocumented” books. Share your findings with other people. But don’t
get mad when we within the MS kernel team have to stay within bounds.
- Jake Oshins
Windows Base Kernel Team
-----Original Message-----
Subject: RE: RIT (Raw Input Thread)
From: “Dan Partelly”
Date: Wed, 19 Jun 2002 09:35:43 +0300
X-Message-Number: 2
I do understand your reasons, as enumerated here. Nor do I need to know
whats a PFN to write a driver.
Yet I try to learn.Because some ppl may be curious about things, simply
because they want to learn how
they are implemented, what particular solution was choosen, etc. Its
pasionating to learn new things about
OS internals, especially when you dedicate a large slice of your time
writting drivers for it.
Hence, I go berserk every time I heard “this is all you need…”.
Best regards, Dan
----- Original Message -----
From: “Doron Holan”
To: “NT Developers Interest List”
Sent: Wednesday, June 19, 2002 3:34 AM
Subject: [ntdev] RE: RIT (Raw Input Thread)
I am not trying to be obtuse, but basing a driver on how the RIT
operates is a recipe for disaster and does not port from release to
release. The algorithm for which the RIT reads from input devices
changes from release to release. Currently the basic algorithm uses a
mizture ZwReadFile, APCs, and a thread (hence the RIT 
).
What is of concern for input drivers is that reads come down fast and
quick and that they are completed as soon as data is available.
D