[Info-vax] Dave Cutler, Prism, DEC, Microsoft, etc.

[John Wallace]

On Nov 11, 8:47 pm, "P. Sture" <paul.nos... at sture.ch> wrote:
> In article <hdajgu$1c8$0... at news.t-online.com>,
>  Michael Kraemer <M.Krae... at gsi.de> wrote:
>
>
>
> > jls schrieb:
>
> > > My recollection from that time period is reading a few articles
> > > written in mags from erstwhile VMS Internals experts that NT had VMS
> > > written all over it.
>
> > > It may not be so true today, but the earlier versions, IIRC, were so
> > > VMS-ish internally that even some of the code was copied verbatim
> > > (i.e., including comments with the initials of VMS engineers).
>
> > If WNT == VMS++
> > then how comes that the allegedly unhackable OS
> > turned into the most hacked one?
> > And how comes that absolutely nothing reminds me of
> > VMS whenever I have to touch a Windoze box?
>
> As I recall the Windows APIs were forced on it. There was also an outcry
> from people who knew about operating systems when it was decided to put
> the graphics in the kernel at NT 4.0.  The motivation behind that was
> allegedly to improve performance with games.

Not just games, but particularly anything which was video intensive.
This was back in the days when a Matrox Millenium was a high end
graphics card.

The originally proposed NT architecture had everything nicely
separated, so there was little risk of total system crash/BSOD, but in
order to achieve this there had to be a lot of parameter copying and
context switching, and consequently there was a performance price to
pay. Performance is trivial to measure with benchmarks. Productivity,
however, was better under NT than the pre-NT OSes because there were
fewer hangs, fewer "out of memory" errors, fewer BSODs, etc. But
productivity isn't easy to measure. So the benchmarks won and
productivity was the loser.

Oddly enough the Trusted Computing stuff in Vista, with its end to end
HD content protection provided by cryptographically signed remotely
revokable drivers with tamper protection so any unauthorised driver
tinkering potentially results in inability to decrypt DRM-protected HD
content [1], sort of reintroduced some of the same inter-subsystem
protections and isolations (but for somewhat different reasons), and
not just in the video subsystem either. And look how well Vista
worked, a decade or more after the drivers initially all got lumped in
to kernel mode.

I haven't yet been able to find out whether this DRM stuff has
vanished from Windows 7.

>
> What irritates me looking back at that time is that NT ws being actively
> promoted by many as being reliable _because_ of its VMS heritage.
>

Indeed. And it could in principle have been that way, if Gates and his
engineering organisation had wanted it to be that way. But it didn't
work out that way, and the rest is history.


[1] Slide 4 in (URL will wrap):
http://download.microsoft.com/download/0/5/0/050a2d04-7432-4325-a5c3-dcbd54cf6695/Kernel%20Mode%20Code%20Signing%20on%20Windows%20Vista%20and%20Windows%20Server%20Longhorn.ppt