[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Device Drivers as Entry Barriers -- Stuart Johnston's comments
Stuart J. Johnston, a senior editor for Information Week, who writes
about operating systems, has has a
"Redmond Watch" Monthly column on Microsoft
(http://techweb.cmp.com/iw/author/redmond.htm), sent me this rather long
but interesting rebuttal to Brett Glass's comments about device drivers.
Jamie
----------------
Subject: RE: Device Drivers as Entry Barriers
Date: Tue, 10 Mar 1998 17:21:32 -0500
From: sjohnsto@CMP.COM
To: love@cptech.org
Jamie:
I took your advice and revised my previous message and included further
information. The revised version appears beneath your note below. I
think that version should be okay to distribute and I'd prefer it not
be
edited further to fit anyone else's point of view. I was there and
documented it at the time and I believe the facts are correct, and I
have no agenda for or against Microsoft -- nor for or against IBM.
Stuart
[snip]
----------------------------------------------------------------------
sjohnsto@CMP.COM wrote:
Jamie:
I can't comment on groups like I20, but since I've covered Microsoft
exclusively as a reporter for three of the main industry newsweeklies
continuously over the past ten years, I have a slightly different view
of a few of the issues discussed in Brett's previous message.
First, even Microsoft has trouble getting hardware developers to write
device drivers because, other than the OS code itself, they are the
hardest to write -- because if one crashes, it typically takes the
entire system down -- even in NT or OS/2. This makes it incredibly hard
to debug, since you can't ordinarily both have a crashed machine and
still have access to all the hardware registers to tell you where the
code was when it died. What is usually done is that one hooks a second
machine up to the first one using a link/add-in card called an
In-circuit Emulator so that when the test machine freezes you can go in
and look at the registers, etc. Writing device drivers is still
expensive and exceptionally time consuming, however.
They are so hard to write that today NT still doesn't have the support
that Windows 95 does because the two device driver models are very
different and the hardware vendors saw the economic opportunity in
Win95, which is a high volume product, but not in NT, which even today
has volumes only about a tenth of Win95. Due to that, Microsoft had to
create the Common Driver Model, which appears for the first time in
Win98 and NT 5, so that a hardware vendor should be able to write a
single driver that will run on both -- at least that's the aim. The
reality may be something altogether different -- no wonder Microsoft
wants to kill off the Windows/DOS kernel that underlies Win95/98. The
lack of "rich" driver support for NT has helped hold up its acceptance
among corporate IT departments for the past several years. That
situation is now improving.
It is the economics of writing device drivers, which noone really
appreciates since its just code and is not the palpable piece which is
the hardware itself, that discourages vendors from producing device
drivers for multiple operating systems.
Secondly, it was not Microsoft's fault that people didn't write device
drivers for OS/2. It was, to a large extent, IBM's fault. During the
first few years of OS/2's existence, IBM, on its own version of OS/2,
only released device drivers for IBM hardware. This was also when they
came up with the idea to make the PC hardware proprietary by introducing
the Micro Channel Architecture in the PS/2, and began charging a license
fee to use MCA to keep out the cloners. MCA also had very tight
manufacturing clearances for the connectors so that it was more
expensive to make add-in cards that worked in it -- thus keeping a lot
of hardware vendors out of that market. At one point, IBM even quit
making PCs that had the industry-standard ISA bus that IBM had
originated because it hoped to change the definition of an
"IBM-compatible PC" to a more proprietary one in order to take back
market share from the clone makers. (The ISA bus still survives and MCA
died instead, illustrating the economic futility of that kind of ploy.)
As I said, for the first couple of years, IBM's OS/2 ONLY came with
device driver support for IBM hardware, (so if you didn't have an IBM
printer, you were out of luck unless your printer had an IBM emulation
mode) and in fact, the earliest versions of OS/2 would NOT install on
hard drives that were not made (or sold) by IBM. (This is not urban
legend, by the way, it happened to me at the time.) I'm not going to
try to read evil intent into that, but it wasn't helpful to IBM's image.
For it's part, Microsoft tried desperately to get hardware vendors to
write device drivers for their version of OS/2 -- a struggle that I
documented at the time -- but it was a long hard fight and driver
support remained mediocre until the OS had been around for five years or
more (remember that OS/2 is more than ten years old now).
No conspiracy theories there. Just plain old economics. The door to
profits was pretty obviously with the volume leader -- Windows, not
OS/2.
At that point, Windows was not the dominant OS, by the way -- MS/IBM's
DOS was -- but it was becoming apparent that Windows was selling while
OS/2 was not. Device drivers are expensive to write. If you're a
hardware vendor, you don't want to have to support more than one
operating system if you can help it -- and two at the maximum. So if
the choice was between Windows and OS/2 or Windows and NT, and you were
a hardware vendor, which choice would you make? Most just chose one --
Windows.
Thirdly, I have been told by people who should know (not at Microsoft)
that it has been a fairly common practice for at least ten years for
hardware vendors to refuse to release at least some of the info required
to write device drivers for their hardware -- especially the popular
hardware -- because that encourages other companies to clone their
hardware and sell "compatible" products at a lower price. So a mouse
vendor might leave out some of the info needed to write a device driver
to work correctly with their mice, somewhat in self defense, simply
because this industry is so aggressively competitive and freewheeling.
Yep, Microsoft was probably guilty of this with their mouse back then,
although I could never prove it.
But it's a widespread practice that did not originate with Microsoft.And
I'm not sure that I agree with Brett that fear of being cloned is
unfounded.
Most PCs I see for sale in stores have "Sound Blaster-compatible" sound
cards -- not the actual item. That's got to undercut sales of the real
thing.
So an issue of protecting your own market -- as the hardware vendors are
doing by limiting access to some "trade secret" information -- ends up
hurting people who want device drivers for other OSes that the hardware
vendors themselves don't want to expend the money or time required to
write device drivers for. It's true but it's also hard to see how to
force the hardware vendors to disclose it -- so it's not really much of
a Microsoft issue. Rather it's a hardware industry issue for virtually
every company that has a proprietary piece of hardware that makes up a
significant part of their revenues. If you force one company to
disclose interface information they consider trade secrets, you have to
do it with everyone and then enforce compliance. That sounds like
government regulation, and even McNealy and Barksdale are against
regulating the entire industry.
Device drivers are a very techy field and they are very expensive to
create -- and they are purely a cost-center activity, unlike the
hardware. (If they don't come with the hardware, it's a paper weight,
but you can't usually charge extra for the device driver since it's
required for the thing to work -- kind of like trying to sell cars
without steering wheels. Noone will pay extra for them.) So the
hardware vendors, as Willie Sutton did, go where the money is . . .
i.e., Windows 95. As I said, Microsoft has had great difficulty getting
hardware vendors to write device drivers for NT and it is only with
increasing volumes, and the promise of a single driver standard, that
more drivers have become available -- helping to make NT more viable and
therefore, more popular.
Note that I say this as the author of the Information Week article that
Jim Barksdale chose to quote from in front of the Senate Judiciary
Committee last week, undercutting what Gates said about including such
features because users demand them, that only 26% of IT decision makers
want the Internet Explorer merged with the Windows user interface.
That was from my story and I also wrote two online columns that
discussed IWeek's survey results as well. (The survey, by the way, was
designed by oursurvey guru, Rusty Weston.)
Stuart
Stuart J. Johnston
senior editor/operating systems
Information Week/Bellevue, Washington
425-641-5980 (office)
sjohnsto@cmp.com
http://techweb.cmp.com/iw/author/redmond.htm
("Redmond Watch" Monthly column on Microsoft)
> ----------
> From: love@cptech.org[SMTP:MIME @INTERNET {love@cptech.org}]
> Sent: Monday, March 09, 1998 8:58 PM
> To: AT-MEMBERS@ABANET.ORG
> Cc: antitrust@essential.org
> Subject: Device Drivers as Entry Barriers
>
>
-------------------------------------------------------------------------
-
> --
> Subject: Re: device drivers
> Date: Mon, 09 Mar 1998 14:47:48 -0700
> From: Brett Glass <brett@lariat.org>
> To: love@cptech.org, Multiple recipients of list AM-INFO
> <am-info@essential.org>
>
> At 03:29 PM 3/9/98 -0500, James Love wrote:
>
>
> >Someone has pointed me toward this and other web pages,
> >
> >http://microsoft.com/hwdev/desinit/wdm.htm
> >
> >and suggested that MS is pushing for proprietary hardware interface
> >standards for next generation hardware, making it more difficult for a
> >competing OS to run on a Wintel machine. Does anyone know much about
> >this? Is this a real problem? Jamie
>
> The conventions discussed on this particular page do not, by
themselves,
> make it difficult for other OSes to work with the hardware. The ONGOING
> problem, however, is that development and maintenance of drivers for
> Microsoft OSes takes so much time, effort, and support that most
> hardware vendors do not believe that they have the resources to support
> other platforms. This is one of the things that caused OS/2 to fade
into
> the background: you simply could not get OS/2 drivers for most
hardware.
>
> This is a problem that's tough to combat once an operating system
> achieves monopoly status. (The percentage of additional sales gained by
> supporting other OSes will always be quite small.) This is one of the
> many underlying facts which refutes Bill Gates' claims that Microsoft's
> monopoly status might quickly disappear: the monpoly is, in fact,
> self-reinforcing.
>
> So far, the alternative operating systems which have the best hardware
> support are -- ironically -- the freely redistributable ones, such as
> the various free implementations of UNIX. In these cases, users
> interested in the hardware donate their time to create the drivers, so
> the hardware maker's investment is almost nil.
>
> Users of these operating systems still run into problems, however, when
> the hardware vendors refuse to reveal the information required to write
> drivers for their products and also refuse to write them themselves.
The
> reasons range from fear of retaliation by Microsoft to fears (usually
> unfounded) that keeping the software/hardware interface secret gives
> them a competitive edge. But since (again) the benefits of having
> support for competitive OSes are perceived to be quite small, even the
> slightest rumbling from Microsoft can cause a hardware vendor to
> withdraw support for driver development. The vendor has much more to
> lose if its drivers are not included with Windows than if it does not
> have support in, say, Linux.
>
> Another problem for the developers of alternative operating systems --
> especially ones for which source code is freely distributed -- is the
> advent of cartels such as I20, which withhold necessary development
> information about whole groups of hardware products from developers of
> alternative operating systems. While I20, in particular, claims that it
> is bound by patent licensing agreements, this claim does not ring true;
> the techniques used by these intelligent peripherals are no different
> than those used by others, and most of the patent claims in question
> would probably be thrown out due to the presence of prior art.
>
> The real challenge, in restoring a competitive environment, is to
> provide real incentives for the support of competition and
disincentives
> for the perpetuation of monopoly. Many in the industry are pessimistic
> about
> the feasibility of any such approach, because it requires long-term
> thinking rather than a focus on the short-term bottom line. They also
> believe that, in the current business climate, Microsoft as a
monopolist
> can easily discourage all such efforts merely by rattling its saber.
> What do you thihk?
>
> --Brett Glass
--
James Love
Consumer Project on Technology
P.O. Box 19367, Washington, DC 20036
love@cptech.org | http://www.cptech.org
202.387.8030, fax 202.234.5176