Perfect Storm Brewing: The Linux Desktop – Part One


Paw Prints: Writings of the maddog

Jul 14, 2012 GMT
Jon maddog Hall

From time to time I get the question of “Why has Linux failed on the Desktop?” Recently Linus was also asked this question, and he considered it a personal failure, since his first desire was to have Linux as a desktop machine. He attributed this to the fact that end user customers just do not like installing operating systems on their machines that they purchased.

I both agree and disagree with Linus.

First of all, Linus, you have not failed. Linux is winning, and will have world domination.

But the world of “consumer retail” is made up of two things:

  • Volume
  • Shelf space

Lesson One: Stores sell what they are comfortable with.

I remember walking into my first computer store, probably around 1984 or so. While this might sound late to some people, I was not interested in “personal computers” at the time, because I had all the computers I needed at my fingertips. Sometimes it seemed as if I had more than I needed, because I had just taken a job with Digital Equipment Corporation, and had joined their Unix group.

The store had three types of computers, Apples, IBM PCs, DEC Rainbows and some low-end PDP-11s. There were three boxes of software on the shelves:

  • Wordstar
  • Supercalc
  • Some modem program

I asked the sales person about the Rainbow. “Oh yes,” he said, “Best personal computer on the market.” He then led me over to the IBM PC and told me all about it.

I asked him later why he did that, if the DEC Rainbow was “The best personal computer on the market.”

He told me that the Apple representative came in once a week to tell him about new things, ask him if he had any questions, etc. The IBM representatives came in every three days. He had not seen the DEC representative for six months.

This weighed upon me for two reasons. One, I worked for DEC and secondly the corporate headquarters for DEC was only about twenty miles from this store. He was in DEC country and could not even get good sales support.

So lesson number one is that the software and hardware can be great, but if the store does not feel comfortable about selling it, then it does not get sold.

Lesson Two: The elbow

Today many people talk about PC computers as “commodities” and “consumer items”, but PCs are really not not consumer items. Consumer items are things like shavers, waffle irons, cigarette lighters, and other things about which you do not ask questions.

When personal computers used to cost 15-20 thousand dollars stores could afford to have well-trained technical sales people answer questions about it. The store hired geeks who were enthusiastic about computers and knew a lot about them. In fact the store was “happy” to have you ask questions the whole day as long as you took the computer home at the end of the day, because the mark-up on the computer price from wholesale to retail generated enough revenue to pay for this customer training session.

But then as prices dropped it got harder and harder to justify answering questions, since the overhead of the sale was eating up profits.

Fortunately (for the stores) as the prices started to drop multimedia started to arrive and people needed CD-ROMs, more memory (for manipulating those digital pictures) and nice sound cards, amplifiers and speakers and the total price rose again to a comfortable level.

As “good-enough” audio became built into the motherboard and the prices started to drop again, 3D and gaming drove the prices back up again.

As the prices started to drop again, large numbers of (high margin) third party applications showed up to drive the final sale price up again, and the management was happy....for a while.

This place in the sales price of the product is called “the elbow”, and as you get near it profits on a per-sale basis start to drop, and management hires less knowledgeable sales people and tells them to spend less time with a customer that has questions.

Now we come to present day. You go to a “big box” store and there are many computers on the shelf, all turned on with their specification cards beside them. At the front of the store is a stack of really big boxes with “total computer systems” in them at really low prices. These are the “loss-leaders”, just like you have with food and other commodities.

In a food store if you have to ask a question about the tangerine you are about to buy, the food store loses money. It is the same with consumer computers.

If you do need real technical help, the really technical people are probably hiding up at the service desk, and you may have to pay for it. This is the path of consumer computers.

Lesson Three: Volume is everything

I have written about this before, so I will not spend a lot of time discussing it here. Volume is what causes commercial companies to port applications to your platform, to write supported device drivers to your platform, to gain a place in the industry. Linux at 2-3% of the installed desktop would probably not be in a very good place if not for the role of Linux in servers, the embedded space, high performance computing and Android; but all of these other markets keep Linux in the forefront.

Lesson Four: Consumer Computer Stores do not sell computers

They sell shelf space. Objects on the shelf need to turn over quickly. If you do not sell an object quickly, you retire it and replace it with another object. Unless the object is a high-margin item, then you may allow it to stay for two days.

Windows has 92% of the desktop market, Apple OS/X has about 6%, Linux has about 2%.  The Linux new sale rate, however is higher, probably around 3-4%, which means that we are gaining over our "rivals".

Now percentage of the desktop market is very hard to calculate for about a zillion (at least) different reasons. Number one is the fact that almost every “PC” desktop has Windows installed, but some people remove Windows, or leave it installed and use Linux instead. Secondly, Software Piracy is rampant, but a lot of that piracy is re-installation of windows on old machines or replacement of “limited functionality versions” of windows (mostly for non-US markets) with fully-functional ones.

But the fact remains that when a customer comes in to buy a PC computer they will typically buy a Microsoft Windows computer nine times out of ten. It simply means that the shelf space holding a Microsoft computer is nine times more valuable than the shelf space holding a Linux computer.

Apple computers, which make up only 6% of the market have a high mark-up, and usually the sale of an Apple computer also sells an Apple Printer, Apple Monitor, Apple Speakers and Apple nose ring....but I digress. The Apple shelf is a higher-margin than the Microsoft PC shelf, but lower volume....inch by inch.

What happens when a Linux person walks in?

First of all, the Linux customer asks a lot of questions about the hardware, wanting to know each and every detail of the specification. They buy a minimum of memory (if they are buying a whole system), configured a specific way so they can go out onto Ebay and buy more at a lower price than the store offers. They also buy a small system disk. In other words, the Linux person is interested in the CPU, the motherboard, and the graphics card, if that.

Then the Linux person argues for half an hour (or more) about why they should not have to pay the “Microsoft Tax”. The store tries to tell them that the “Micro$oft Tax” was paid by the motherboard manufacturer, the license passed through to the OEM and then passed on to the store OR the license was paid by the OEM and bundled in, so the retail store really has no idea how much money is paid in the “Tax”, ergo do not know how much to refund or if the store will ever get reimbursed for this money.

But remember, this is half an hour of discussion...much too much time for a profitable sale at the store, even if they do not give you back the “Micro$oft Tax”.

Then, after all that sales help, most Linux people do not buy anything else.

In the early/middle days of GNU/Linux, you might find boxes of Red Hat Software, or SuSE or some other flavor of distribution (or perhaps all three) on the shelves of a computer store.

Some Linux people (I am one of them) might buy the box to get the printed documentation inside, or because their Internet connection was slow, or even just to give support to the distribution and to reward the store for even considering selling the operating system.

After a while, however, the store owners recognized that boxes of distributions had a very short shelf life. In fact, according to their customers by the time the boxes got in the store and on the shelf the distribution was “old and crufty”. In the box was Red Hat version 5.2, and 5.2.1 was already on the Internet!

So the boxes tended to stay on the shelf, and get older. Even if the distributions were willing to take returns of unsold boxes, this too was expensive for the stores, since they had to do the work of returning them, keeping track of returns, and of course not making any money on the shelf space.

Even worse than that, Free Software people did not buy any “layered products”. No Adobe Photoshop was purchased and the FOSS people kept talking about “GIMP”. When the store owners would ask where they could purchase GIMP software from, the FOSS people would just laugh and tell them “it came with the distribution.” Another major source of income lost for the store owner.

There was one source of income that was better with FOSS than with closed source products, and that was integration and customization of the FOSS software. Unfortunately this was an income source that consumer systems typically did not need, and store owners did not have the expertise to deliver. This income stream was seen by large OEM companies like IBM, HP and others into commercial accounts. These OEM companies typically did not buy their computers from the “Big Box” consumer stores, or from the mom-and-pop stores.

There is another factor that enters here. At least before the advent of tablets and notebooks, in the days of the “Desktop PC”, about 60 percent of all the desktop PCs did not come from IBM, Dell, Apple, HP, the “large PC vendors”. Sixty percent of the desktop computers came from “white box” companies and stores. People who would take a case, a motherboard, a graphic board, etc. and put together a made-to-order system for use by a company. Or perhaps a hardened system for military use or use in a dirty, dusty environment. Or perhaps for use in a wet, humid environment. We called them “white box” because they did not bear the color or logo of some large company.

These “white box” integrators might be a corner PC store or they could be a company that specializes in meeting the needs of military, or industrial needs. They typically have more expertise than the “Big Box” store, but their expertise is in Micro$oft products, not FOSS.

One other issue is the cost of a product line. At Digital, every time you introduced a new product, it might cost 20,000 dollars just to do the administrative paperwork for that product. You had to get a name and order number for the product, put the product description into the catalog, put an entry (or two or three) into the price book, the process went on and on. You then did training of the sales force on the new product, the support people how to support it, create marketing materials for the customers, etc. etc. Return forms, Product Descriptions, Warranties and Service Level Agreements, all have to be finished.

Much of this still goes on in large companies, which creates “overhead”.

When people say “Why not just create a GNU/Linux computer?”, to many large companies this means a lot of money for bureaucracy, and (perhaps) very few sales.

Some of these things are changing slowly over time, and the Internet is both helping and hurting this.

On the Internet it is really hard to run out of “shelf space”, so companies that sell over the Internet can afford to have a few systems that have Linux installed. Sales over the Internet can be “drop shipped”, installing Linux on demand, so the version of Linux can be easily kept up to date as well as well as reducing Linux-specific inventory. Also technical questions tend to be treated more with forums and FAQs.

On the other hand the Internet also tends to drive prices (and profits) down, so competition is harder, and particularly when you do not add much value in the way of technical support.

But there is a “perfect storm” that is brewing, and because of that storm, things may change rapidly.

Stay tuned for part two of this blog entry, due in a couple of days, which will describe that “perfect storm”.


  • Linux and "White Box" Computers

    I agree with most of what the author of the article has said, except with respect to Linux penetration of the desktop which I have good reason to believe is very much higher, especially in Asia. Nontheless, the description of the economics of retail shelf space is still true. With high real-estate prices the value of shelf space is even higher and volumes and margins must be high or the store is doomed. Also, many stores receive "push money" to put items on the shelves. In other words, manufacturers literally pay stores for shelf space. They do this for groceries and they do it for IT.

    "White Box" suppliers do not face the same pressures and are generally closer to their customers - which are generally other businesses. It is my experience that over half these systems are supplied with one form of Linux or another. Frequently the boxes are set up to only run ONE single specific application, which is often custom. Examples are Point of Sale (POS) devices, teller terminals, and logistics control applications. Likewise most industrial controllers are Linux based. When you add it all up, it accounts for a staggering number of boxen.

  • Still a Pipe Dream

    Interesting article and many valid points. Having attempted to go at least 90% Linux from 2005 to 2010, I finally gave up in desperation. Not that I'm a heavy gamer or in need of specialized codecs, it was just a lack of consistency and usability. ( - for an example of the work needed to get a simple software to work in Linux.)

    Though Windows is seriously flawed - right down to the continued reliance on drive letters - it is at least consistently flawed and the applicaitons likewise. Simple tasks under most Linux distributions require deep knolwedge of command lines. In addition, configuration in one distro may be vastly different that that in another (YaST vs. Sinaptic, for example).

    I do like how the Android ecosystem is taking off. I find myself using my Android-based tablet and smartphones more often than my home desktop. In that manner, i can say that I use Linux still on a daily basis. I also have Linux installed on an older laptop as well as on a USB key for emergencies.

    I think the year of the Linux desktop was maybe 2005/2006. But then not enough happened to make Linux fit with the modern digital rights era and eventually sofware companies passed it by.
  • Much boils down to support

    I started with desktop micros in 1977. The first adopters were small businesses and what they demanded was support. Be it educational, technical or service, the deciding factor was most commonly the support the vendor could provide, not the technical details of systems or software.

    The great success of the IBM PC in business was in no small part due to the fact IBM had legendary support. Burroughs, DEC, Wang and a host of other competitors were out there, but no one had the reputation for support of IBM.

    You mention comfort as the factor that led to stores stocking IBM, but the anecdote you cite shows that marketing and technical support was what led to that confidence. With the others, you were on your own; with IBM help was a phone call away.

    With the exception of Red Hat, few Linux distros have enjoyed much direct support of any kind. Without trainers and technicians Linux and other Open Source software suffers a huge disadvantage in the marketplace. Desktop Linux is a product embraced by the technical cognoscenti, but without clear avenues to obtain the answers to questions and solutions to problems, few others can be expected to make the leap of faith.
  • Another aspect - drivers

    I am a reasonably knowledgeable Linux fan, but by no means an expert. 6 months ago I bought MoBo GIGABYTE GA-880GM-USB3 R and the ASUS Xonar Essence ST sound card. Installed SUSE Enterprise. Could not get a mobo driver for Linux. Could not get a sound card driver. Contacted ASUS. "There is NO Linux driver for it," was the emphatic answer from the ASUS engineer on the other side. I could only get Windows drivers for both. I understand that ASUS may have an exclusive deal with M$oft, but if the vast Linux community had provided drivers for these popular pieces of hardware, I would be a happy Linux user today. Since I was not able to find the mobo and sound card drivers for two months, I did what I swore never to do - I uninstalled Linux and installed W7. My comment/question is: I think there are quite a few people like me - not experts, but willing to install Linux and put in some effort into it. Wouldn't these 5 per cent increase significantly if we could readily get the necessary drivers?
  • Good stuff!

    I like where this is going, and look forward to the next entry.

    In addition to the other reasons you noted, perhaps the biggest problem with the ever-falling margins for PCs is the refusal of the manufacturers to differentiate.

    No matter how much they spend on R&D, now matter how much they explore different configurations, in the end all of that is a waste of money the very moment the user turns the machine on. At that moment, the user experience is handed over to Microsoft, and since most PCs makers ship the same OS they're delivering the exact same user experience, down to the pixel.

    Using Ubuntu may be one way to differentiate, and even if the OEM cost of Windows is low these days, not having to pay it would likely double the slender margins PC makers have today.

    But I'd go even further to suggest that PCs with Ubuntu preinstalled could sell better if priced slightly higher than Windows machines. It's price positioning: up to a point you can engender a product with a higher perceived value by raising its price.

    I wouldn't advocate attempting Apple-like prices, averaging twice that of PCs. But imagine a small number, like $10 or even $20, over the same machine with Windows installed. That's not enough to stop the sale, but will communicate the additional value of not having the same-old same-old.

    Think REALLY different. blunk
  • Todays event is a huge leap forward for the Linux Desktop

    Valve bringing steam to Linux is sure to help the acceptance of the Linux desktop
  • It's not about shelf space

    Elaborate explanations about retail shelf space are unnecessary to explain why Linux has failed on the desktop,

    As a desktop OS, Linux fails, has failed, and will continue to fail for one simple reason: The user experience SUCKS. Nothing more complicated than that.
comments powered by Disqus