Peter:
This is not exactly hard to believe. I’m just not sure how much size
matters here, as it were. That is, I can’t really think of any reason
why I would chose a smaller package (ibid) over being able to get one
package and know that I was good to go, I would be mighty unhappy in
many cases if a network connection were required, and finding out what
you’re missing only during install, unless you happen to know the
contents of every machine on which you will be installing, while you
read carefully the documentation for the driver installation, that must
be accurate, is the worst case, for me, at least.
Hagen:
This is absurd:
People have become used to have separate drivers for Win16, Win32s,
>Win95, Win98, Win98SE, WinME, Win2000, Win2003Server, WinXP and
>WinVista. Oh, and the 64 bit versions, of course.
>So used that some of them who happen to be driver developers even deny
>the possibility of writing one driver for everything from Win98 (with
>limitations) up to Vista. 
I gather you have the misfortune of supporting Win98, as you bring it up
a lot, if I am remembering correctly. You’re one of the few here that
does. Perhaps it’s possible that people merely don’t talk about Win98,
because they don’t care. Also, of the ten versions of Windows you
mentioned:
-
Win16 is pretty clearly not something has any expectations about
either way at this point.
-
Win32s is something no one ever had any thoughts about at any point,
and drivers really don’t have much to do with it.
-
While there are some differences between all the Win95 family, not
all that many people are concerned with any of these at this point, and
I having trouble with the idea of anyone expecting a different driver
for Win98 for Win98SE, at a minimum.
-
You left of WinNT 3.51 & 4.0. There’s no reason to discuss them, as
I can recall exactly one question on this list in three years about 4.0,
but if you’re going to talk about Win32s, why not these, and the Itanium
for that matter as well.
-
Win2000 is still out there, but a lot of drivers that work on XP
work in 2K as well.
-
Win2K3 supports almost everything that XP does.
-
XP is what the only thing that most people care about at the moment.
-
Vista is indeed different, and those who use it no doubt expect a
separate driver package.
-
Also certainly true for 64-bit.
What I really don’t understand is given how outrageously awful
installation has always been, easily the worst feature of windows, and
that this is what causes a lot of issues today, and will continue to do
so tomorrow, why one earth would focus on driver package size. How does
this affect you? I mean, sure it takes longer to download, and if we
were talking about why is the WDK like 2GB sometimes because of DTM, or
that you have to download about 6 GB to create a partial checked build
(unless you already have imagex), I would heartily agree with you, and
add that the whole thing is made far worse by FTM, and that in some
cases if you don’t know to burn a lower speed, the OS iso’s don’t really
work. But how big can a driver really be? Bigger than necessary, sure,
but it doesn’t even chart.
mm
Peter Wieland wrote:
As long as we’re relying on magical things we’ve imagined that don’t exist, we might as well have the computer call up and have a CD with WDF delivered to the user’s home in a golden carriage pulled by badgers wearing little white gloves and top hats. Then the badgers can install it for you, give you a quick mink massage and go on their way without even leaving footprints. Unfortunately no such feature exists in previous versions of Windows (and I think the legal risks associated with the badger delivery service would be too high for Win7) so we’re left with reality.
The input we have gotten around setup was overwhelmingly that driver package had to be able to carry its dependencies with it. It couldn’t rely on a service pack or WU update being on the machine. It couldn’t rely on them coming down over the network. It has to work without any user interaction or UI.
We have the resources to put together *one* setup method … this was the most requested scenario and functions in the other cases and so it’s the one we implemented. We have some ideas about how to streamline the installation but it’s not clear when we’ll be able to make those bubble up to the top of the list of all the things we’re being asked to do (from external and internal partners).
Don’t worry about the rant. Clearly it was my mistake for trying to answer your question. My apologies for sending you off the deep end.
-p
-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of Hagen Patzke
Sent: Wednesday, February 27, 2008 12:40 PM
To: Windows System Software Devs Interest List
Subject: Re:[ntdev] Handle Invalid error when trying to run a KMDF driver
Peter Wieland wrote:
> It’s a tempting idea … the problem is that it then ties all this
> new device installation technology into specific versions of Windows.
> Sure your driver package could state that it needs update Y but the
> code to understand that is probably in update X which has to be
> applied first.
So you are telling me that in all of these Windows older driver
installers there is absolutely no mechanism to inform anybody that a
system component necessary for driver installation is missing?
Not even with a text message?
Come on, even then there could be a ‘stub’ that informs the user of the
fact. As we had (still have?) in Windows EXE files that tell you “This
program does not run in DOS mode.”.
And don’t underestimate the amount of install hassle ordinary Windows
users have become used to during all these years of “Windows
Experience”. YOu use an outdated version? OK, it will even be harder.
Or someone has to get a system architect to MS. Quick. Swap her for
someone from marketing - they are excellent and surely can spare one person.
> We’ve made the assumption that no one wants to have separate
> installation packages - one for pre Win7 and another for Win7 to deal
> with a radical change in the installation model. So we instead have
> to use a coinstaller.
People have become used to have separate drivers for Win16, Win32s,
Win95, Win98, Win98SE, WinME, Win2000, Win2003Server, WinXP and
WinVista. Oh, and the 64 bit versions, of course.
So used that some of them who happen to be driver developers even deny
the possibility of writing one driver for everything from Win98 (with
limitations) up to Vista.
Great - I’m looking forward to a brave new world where every driver
package will even bigger than in earlier times because now not only a
bloated install program is delivered (to make sure the files are
actually found), but additionally with every WDF driver comes a
coinstaller. Well, the price of progress, I guess…
Sorry for ranting, but this reminds me painfully of a three-disk
installer for an external CD writer that would not run on NT4 with a
“file not found” message.
Luckily it was an old archive type that I managed to find an unpacker
for. Turns out that more 2 and 3/4 disks were actually the (compressed!)
install program.
The real driver consisted of one INF and two SYS files, totalling less
than 200kB. And of course someone had made a typo in the INF file, and
obviously forgot to test it on NT4.
The standard Windows device installer was absolutely happy taking the
INF/SYS from a subdirectory on a single 3.5" disk.
Now we will at least a single CD. Or DVD. Heck, now I know why BlueRay
had to be invented. (Grrrrrr!)
So much my 2cents of steam-off. No offense.
NTDEV is sponsored by OSR
For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars
To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer
)