Linux does have this and it is better than the NT implementaion. If
you mmap the file, the kernel doesn't need to copy before sending it
over the wire. There was a thread on this list about it.
>
> As is then current NT web servers actually perform very well.
Not as good as Apache. On the same hardware, Apache runs better under
heavy load. The tests that reported differently were not executed by
compentent test engineers. More to the point, however, is that the
features of Apache overshadow all other Web servers. The real
problems are in generating dynamic content, not in serving static
pages.
> Bill's O/Ses have success because software/hardware companies trust M$$$
> perennity. The main enhancements of NT have been to add to Bill's
> Windows System features that existed since more than 15 years on Unix
> and proprietary systems.
>
> There is also a degree of certainty about that will be available to end
> users a year down the track. Still, things like WinG do happen.
>
> If a vendor writes a device driver for Windows95 - generally it will work
> with a newer kernel. The same cannot be said for linux. Things like the
> module API and stuff are too volatile/sensitive to changes.
This is, unfortunately, very untrue. Versions 3.5 and 4.0 of NT have
very different driver architectures. My last driver required three
DLLs because of the funky *new* design. In porting linux drivers,
there is documentation and on-the-net collaboration. The support
engineer at Microsoft, after several hours of phone time, could never
answer my questions. A couple of weeks later, he called *me* to glean
my answer because other developers were calling about the same thing.
-- Oscar