Mail Archives: cygwin/1996/12/06/21:28:52
[Big foot in mouth coming, pinch your nose]
I spoke out of temper, without thinking, and I said some stupid things.
Apologies to all, especially those who really dont give a damn about
this thread
[like myself].
There are two sets of APIs captured by those two DLLs I indicated: the
user-mode,
Win32 APIs, and the kernel-mode APIs. The former are documented in
the VC++ Win32 help files, the latter only in the DDKs (DDK, HAL, IFS).
The NT group's policy is that unless it is documented in an SDKs or a
DDKs they
dont feel like committing to a certain API. The only officially
supported user API sets
are Win32, OS/2, and POSIX. As you realize, all of these APIs have
external constraints
that make them non-NT-specific. The company does not want to give birth
to yet
another kernel interface, therefore there are no official NT-specific
user APIs.
The documented NT-specific APIs are for device drivers, running in
kernel mode,
and even those will become common to NT and Win95 (with the WDM model).
As someone else pointed out, cgwin32 should rely only on the Win32 APIs
(because they are) supported on NT *and* Win95. The OpenNT thing is
targeting NT exclusively [and I dont know nothing about it except press
releases].
The article you pointed out contains noise and one data point: the
"cancel I/O" thing.
What they meant was that in kernel mode (e.g. it wont help you) it is
possible to cancel
an IRP (I/O Request Packet), but that device driver writers are strongly
discouraged
from doing this. [And since I have written device drivers myself, I
really dont see why
you should get in the woods of using such approach (just return the IRP
with error, right ?).]
Now I swear big time, no matter what you do or say I am not going to
speak on this again.
sandro-
>
-
For help on using this list, send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".
- Raw text -