Robert Pogson

I have always been amazed at how much memory XP needed to amuse one user. I can amuse sixty users in 4 gB RAM. “7″ needs that for one user… Wow! For a lean, mean OS, “7″ is a resource hog. I guess the 600 million XP PCs out there will not be migrating to “7″. They could switch to GNU/Linux instead of staying with obsolete XP.

Seriously. My people are running XP in 256MB with a serious over-commit to virtual memory. I can run 12 users at once and services in 1024 MB with GNU/Linux. That would explain a lot of the speed difference. My terminal server is not swapping.
Update: SJVN opines that “7″ is not a resource hog, but he does say, “Now, that’s not to say that Windows 7 doesn’t require a lot of memory real estate. It does. I recommend a minimum of 2GBs of RAM for 32-bit Windows 7 and 4GBs for 64-bit Windows 7. For choice, I like to give either version at least 6GBs.”

Ah,,, resource hogs are not resource hogs when you give them enough slop… . At least some good comes out of that other OS. The price of big RAM keeps going down because they keep thinking if they just produce more they could make some money, like all of M$’s partners.

Tesseract is a fine FLOSS OCR (Optical Character Recognition) package that can help us expose the truth about M$ to the public through Google. Google indexes text pages very well but many of the .PDF documents in records of courts are images of documents that have been photocopied a few times too often. Tesseract can convert them to text.

Groklaw has been doing a fine job of this in Comes v M$. I thought I would spread some joy to US DOJ v M$.

Here is an example: http://www.usdoj.gov/atr/cases/exhibits/365.pdf.
I convert it to TIFF images and turn Tesseract loose on it to get text. Debian has all the packages I need. Here is the result:

“$ls

365.pdf

$pdf2tif  365.pdf

$ls
365-01.tif  365-02.tif  365-03.tif  365-04.tif  365.pdf

for  f  in *.tif; do  tesseract  $f  $f; done
Tesseract Open Source OCR Engine
Tesseract Open Source OCR Engine
Tesseract Open Source OCR Engine
Tesseract Open Source OCR Engine
$ ls
365-01.tif  365-02.tif  365-03.tif  365-04.tif  365.pdf  365-01.tif.txt  365-02.tif.txt  365-03.tif.txt  365-04.tif.txt

cat *.txt  > 365.txt”

Here is the text after fixing glitches:

From: Joachim Kempin

Sent: Tuesday, December 16, 1997 5:37 PM

To: Bill Gates

Cc: Steve Ballmer; Paul Maritz; Joachim Kempin

Subject: As promissed OEM pricing thoughts

draft OEM DTOS

PRICING.doc feedback appreciated

MS7 007193

CONFIDENTIAL

DT OS pricing strategy

During our offsite last weekend the OEM team discussed this issue and this is a summary of our

conclusions.

Current situation

The current ASAP decreases for PC manufacturers will make us a much higher component of their system cost then ever before. We expect that <1k PCs will be bought by consumers and business and could constitute more then 50% of all PCs by C-mas of 1998. In case we see $500 PCs be next C-mas our royalties we could be as high as 10% of total system prices and if the biz PC markets gets eroded by <1k PCs we will with an NTWN solution be in the same position. While we have increased our prices over the last 10 years other component prices have come down and continue to come down. This is in particular true for CPU prices. where AMD and CYRIX are clearly under $50/unit components with packaging COGs of $20-25. Intel has higher costs today because of their packaging and I estimate that their current average CPU price is around 170-180$ with 40-60 $ in packaging costs (so the money they are getting for their IP on silicon is $120-140 in average, which compares with NTW prices being between $100 and 120 typically). I am interested in listening to them when they explain to us their low-end strategy in Dec. When comparing system prices over the last couple of years with today’s prices we should note that in the low end segment PC manufacturers have started pulling out monitors and other items from their systems. As a result my comparison is not 100% correct- but this does not change the trend. We have to assume that not all of the <1k PCs will be less powerful or just consumer focussed. Many will be less capable as OEMs strip peripherals. We are making this easy with USB, 1394, device bay etc. Easy transfer of peripherals to new PCs could be the result and the vanilla core architecture might get artificially even cheaper. At least this would make it easier for us to defend our pricing as we increase our BOM percentage.

OEM division revenue growth over the last 8 years has depended heavily on volume increases and a trend to higher priced OS. During that time ASPs have stayed stable or have gone up which made it easier to ride the wave and get the value we deserve. We have shown larger then 40% growth rates annually and expect in the future that OEMs will take a very hard look in how to avoid paying us more $5 per system in order to hit most aggressive price points. Will this lead to significant higher volumes and thus allow us to relax some prices while gaining share where we need it? The danger does exist that more PCs might get shipped without an OS and we should not take this lightly!

While reasons for volume increases are too early to analyze (US data still sketchy and ASIA/LATIN data really convoluted) we expect the following to happen:

1. Moderately more volume by Ending new buyers who can now afford to buy PCs

(This should be true for consumers as well as small biz)

2. Acceleration of replacement cycles

(Knowing that 80M PCs cannot run NTW or WIN 98)

3. Shortening of PC “life time” in general

The only counter argument to make here is that current PC technology is totally sufficient for most office tasks and consumer desires and that any performance bottleneck is not in today’s PCs but in today’s COM pipes. This in itself might slow down replacement cycles and life time shortening until we find true MIPS eating applications- a priority not only INTEL should subscribe to.

Other side effects of the <1k PCs are less need for NCs, NetPCs and WIN terminals as long as we deliver on the well managed aspect of the PC environment within 12 month. If not customers might not wait for us and pilot more alternative solutions. I do not have to say what this means for NT 5.0 delivery.

MS7 007194

CONFIDENTIAL

Pricing options .

PC industry growth after the Asian crisis settles down should go back to normal and might wind up for CY 99 and 2000 in the 20%+ figures. This could help us to ease up on increasing prices – but the drive to N’l'W needs to continue and as we go along we might conclude that the market will not bear $100+ prices for NTW. Our options,

1. Peg DT OS prices to type of CPU or system price

Both methods are an administrative nightmare for the OEMs and us. This worked when we had only 3 CPU types and the one with the higher royalty had a long cycle time ~ today we have too many types (I can just Intel calling me feeling we treat them unfairly by putting all their competitors into the low end bucket) and the cycle times are so unpredictable that we recommend against this. We have priced once on manufacturer cost and it is a sure way to totally erode your model without having any control. We rejected this as well.

2. License for limited time and create annuity business

This is the best thing long term but it might disrupt end user operations and could require enduser registration. I wrote a memo about this more then a year ago. This will need technology and infrastructure to be set up something we are not seriously working on. So until NTW 6.0 comes out- say CY 2001 this is not an option. We need a champion for this now, if we want to do this.

3. Reduce DT OS content which OEMs install and sell add on retail packs

This is a viable option if we can make the add-on pack a stunning piece of technology and a “must have” for every PC owner. Performance, management and ease of use features come into my mind. Again we need to start this now in order to be ready at NT 6.0 time frame.

4. Defend current model

We believe that we ducked the bullet for 15-18 month and bought some time to explore the above opportunities. Only 3 major contracts are not agreed upon. The one company who is pushing the hardest for a price break for the sub 1k category is Compaq and I expect a major fight and escalation on this subject. The answer here has to be “no” for all people involved.

With this in mind Iet’s agree on the following objective, strategy and tactics:

Objective:

To get the highest amount of $/unit for DTOS through the OEM channel without breaking the current model of pre-installing the SW on PCs.

Strategy:

Avoid price increases for DT-OS over next 2-3 years and be sensitive to NT pricing and prepared to revisit as we go along.

Tactics:

- Reduce some of the more rigid licensing requirements, which increase costs to the OEMs.

- Step up our marketing efforts with OEMs to help them to sell more PC units

- Give OEMs air cover by promoting high-end PCs purchases by providing more future technology directions

- Continue to level the playing held between SB, Named and MN accounts.

- Increase demand creation for NTW PCs slowing down OEM’s ASP erosion

- Resist <1k PC royalty price decreases firmly

- Reward OEMs who are willing to increase their NTW penetration until NTW 5.0 ships

- Review MOLP and SELECT waterfall as well as Academic and special government pricing options and agree on a company wide pricing. model without allowing any exceptions on subsidiary or area level

MS7 007195

CONFIDENTIAL

Who can derail this plan and MSFT counter tactics:

OS competitors

SUN

Sun and it’s coalition with Java. For the next 2-3 years the barriers are huge for them and even IBM after studying this technology is not convinced it would satisfy customers when implemented during that time frame. In addition there is the compatibility barrier and the fact that OEMs see SUN as the enemy and will not be easily convinced to be a distribution channel for them.

OEM coalition

Our high prices could get a single OEM( Compaq might pay us 750M$ next year) or a coalition to fund a competing effort (say in India). While this possibility exists I consider it doubtful even if they get a product out that they can market it successfully, leapfrog us and would not deviate from their own standard to differentiate. Could they convince customer to change their computing platform is the real questions. The existing investments in training, infrastructure and applications in windows computing are huge and will create a lot of inertia.

No bundling of OS on low end systems would be the easiest way to hurt us- but who would want to start with this and loose business’?

ISV

NSCP may come from the browser side, but I consider them too weak to succeed alone- so they are only dangerous if they team up with SUN. Again compatibility and yet another platform are the biggest inhibitors.

INTEL

We read about it in the news today and over the last couple of weeks. If they decide to own the OS as well as the CPU our business it will get ugly. This could be an INTEL lead and funded coalition- say with Compaq and NSCP. I am convinced they have been thinking about this for some time. They could buy SUN SOFT or start a skunk work project on their own. lf they decide to sell the OS for $1 and the CPU for $ 200 they will get the OEMs on their side. The customer inertia argument remains and that will prevent them to build momentum easily. Our reaction could be to buy Nsemi or AMD or both and own the CPU and the SW business- while both stocks (INTEL and MSFT) are taking a dive. We would sell SW at $100 and CPUs at costs + $1.

How sure are we of our partnership and how fast could we react if needed? We could bring compatibility to another platform better then anybody else and we would have the money to fund the fab capacity.

Bill, please send me some feedback, does it make sense to discuss this with a larger

audience’?

MS7 007196

CONFIDENTIAL

Isn’t that interesting? While M$ was telling the world it was innovative, it was looking at ways to stifle the competition and to sell consumers stuff they did not need. Note that M$ felt pressure from sub $1000 PCs in 1997. How must they be sweating with PCs at $100-$300 and M$ have raised prices again? Well, the higher-priced units are not selling. The lower-priced units are having to cut prices to compete. That means the cash cow is drying up. Do you really need a quad-core CPU and video card that can do 200 frames per second 3D? Do you see competition in the market or do you see OEMs, Intel and M$ colluding to keep prices high? This year, OEMs will be under a lot of pressure to dump M$ because ARM will sell and do it all without the OEMs and without M$. To keep moving units, OEMs will have to cut prices. Hardware is already at rock-bottom, so the cut will have to come in software. Good-bye M$. Hello GNU/Linux.

Business basically buys no-OS PCs in bulk and writes disc images to them or uses thin client technology. They really have no need for the M$ tax to raise their cost of acquisition. Businesses compete. If the competition adopts GNU/Linux and thin clients, others will follow. It’s happening.

They were a little ahead of their time because Java was new and networks and servers dragged but people still call thin clients dumb terminals because of the FUD spread by M$ and its partners. You can see part of the story in these summaries of M$’s campaigns about the time of Lose ’98. That is a document produced as evidence in Comes v M$ and it shows that not only did the NC have a few problems of its own, M$ actively connived with its partners to dig a hole and bury the NC. I will use that document in a grade 9 class outlining the history of the PC. The students will be using network computers/thin clients so they will know FUD when they see it. These are not dumb terminals or Java-based thin clients but regular X terminals showing the pictures and sending the clicks to a powerful machine built four years after the FUD campaign. The NC works. 10% of PCs are NCs today and the number is increasing steadily. They last so long a low level of production is sufficient to meet the needs. My students see the NC/thin client as a smart way to do IT and one that is superior to the thick clients they see at home and at school. Everyone who sees them here wants one.

Anyway, here are the highlights of that document wrt NCs:

• Begin phase one of the Windows Vision: Natural Computing, with above two launches.

• Continue worldwide efforts to prevent the NC from gaining any critical mass.

Cute, eh? Produce a product with a similar name while putting down the other guy’s stuff.

• NC Attack: What we’ve done

• Account visits got us back in the race at FedEx and St. of Florida, whereNetPC’s have been added to NC evals. Visit to AARP resulted in no further NC deployments.
• Which to Choose document, TCO slides, NetPC demo units, and TCO calculator
• NC/Java competitive session #1 rated session at MGS. Delivered tech sessions on NC, Hydra, and ZAW
• Launch NCFacts.com with WagEd in August to deliver low road NC/Java messages. Continue to use ms.com to deliver hi gh road messages .
• Deliver NC trial/rejector case studies (One by end of Sept)
• Sun and Oracle plan to ship their NC s in volume in the fall, we must continue to utilize the press, events and partners to get the word our that the NC’s and JavaOS are the worst of both worlds.
• “There is next to no value in a terminal device. The value is in the content. The [NC] market eventually will resemble the razor-blade market. The devices will became handles and the content, the razor blades” Brian Murphy, The Yankee Group
• NC is likely a long term thrust, despite initial success over the last 6 months.

So, M$ waged a serious campaign to discourage production and purchase of network computers for more than a year while ramping up production of competing technology with Citrix and others. It takes great salesmen to claim a product is superior to another which effectively do the same thing, provide a virtual desktop. This campaign was so effective that many producers of thin clients went out of business at the same time that M$’s monopoly solidified. What’s wrong with this? Good salsemanship? No! This is sabotaging other legitimate businesses and is illegal in US law and other laws around the world. That they got away with it and experienced ten years of monopoly as a result is one of the crimes of the century.

NC was a trademark of Oracle and M$ was deliberately attacking the trademark telling millions of people it would never fly.

The Lanham Act expressly forbids such activity.

(1)

Any person who, on or in connection with any goods or services, or any container for goods, uses in commerce any word, term, name, symbol, or device, or any combination thereof, or any false designation of origin, false or misleading description of fact, or false or misleading representation of fact, which–

(A)

is likely to cause confusion, or to cause mistake, or to deceive as to the affiliation, connection, or association of such person with another person, or as to the origin, sponsorship, or approval of his or her goods, services, or commercial activities by another person, or

(B)

in commercial advertising or promotion, misrepresents the nature, characteristics, qualities, or geographic origin of his or her or another person’s goods, services, or commercial activities, shall be liable in a civil action by any person who believes that he or she is or is likely to be damaged by such act.

…

(4)

The following shall not be actionable under this section:

(A)

Fair use of a famous mark by another person in comparative commercial advertising or promotion to identify the competing goods or services of the owner of the famous mark.

(B)

Noncommercial use of a mark.

(C)

All forms of news reporting and news commentary.

So M$ went way beyond fair use/comparisons. Otherwise why did they suck in the press and the “low road” NCFacts.com? They knew they were doing wrong but took the chance they would not be caught. The crime paid off handsomely. That’s why they continue to push the boundaries of anti-competition law.

In spite of their FUD, the NC is doing well, ten years later. It’s working in my schools. Novell, RedHat, and IBM are mass-producing systems of them with management tools orders of magnitude more efficient than what can be done with thick clients. The netbook, the smartbook, and the virtual desktop are all here to stay and taking share from the monopoly at last. M$ has tried to wring “value” from them by CALs and other horrors such as limiting the hardware on which their OS may be installed and they have tricked most OEMs into not pushing too hard but there is light at the end of the tunnel for the NC and it is not an on-coming train.

Wyse and IDC report huge savings using thin clients. Being partners of M$, and describing large business use of IT they refer to that other OS and not GNU/Linux but the savings will be larger with GNU/Linux just by cutting out the licensing fees. M$ charges server licence, CAL and application licence… It all adds up. In my installations, the cost of licences would have been about equal to the cost of hardware so we installed seats for half the cost using GNU/Linux. GNU/Linux gives better performance too because the server can run twice as many processes in a given quantity of RAM because of shared memory between executables. Others report similar results.