Pre-installing Linux

Recently Dell has been in the news over the top suggestion on the new Dell IdeaStorm site: Pre-installing Linux on retail machines. Today Mark Shuttleworth wrote an article on the difficulties/reasons Dell, and other resellers, may not want to Pre-install Linux. I feel a serious issue has been overlooked in this entire saga, more on that later. Mr. Shuttleworth’s comments are certainly valid, but there are solutions:

  • People who are buying retail, from a reseller, will go for the cheapest solution. If they don’t know about Linux, they are likely to have a problem in the future. Solution: Separate the concept of buying a computer and paying for a company to configure your computer with software. Let’s be realistic, if you are in this category, you are really paying Dell to configure your operating system, you are not paying them to sell you one.
  • Microsoft contributions and advertising offset the low margin on hardware. Solution: Following from the above; customers purchasing a computer configuration, currently, pay Dell for the software, and pay for it too be configured. I see no reason why Dell can not add a similar cost for Pre-installing Linux (i.e. configuring the computer). Customers may need to be educated about the differences between purchasing software (licensing really) and purchasing a configuration process. There would be other issues, but they could be worked through.
  • Fussy customers. If you are going to be that fussy, well, why don’t you order the components you want and DIY. Recent revisions to business models clearly indicate companies perform better with a smaller scope, especially if they stick to that scope, and call alienated customers an acceptable loss, they were never the target anyway. As Mr. Shuttleworth pointed out System76, and others, provide a high level of specialization, let them deal with the fussy customers.

In all of this, there is a serious issue that has been overlooked.

Firstly, consider the shaped-peg into different-shaped-hole analogy where: Windows is a pentagonal (5 sided) peg, OS X is a triangular peg, Linux is a lump of play-dough, and the market is a round hole. What happens? All pegs fit in the hole, some fill it better than others. Generally, the customers who the, non-filling, pegs do not cover, know who they are, and deal with the situation appropriately. Linux produces a problem in this environment, and Mr. Shuttleworth knows it. At the moment resellers have no choice but to make a regular polygon-shaped peg out of a Linux distribution, and push it into the hole. Problem: Linux users want an non-regular polygon-shaped peg, something that is specific to them. Sometimes the regular polygon peg will cover the same area as the non-regular polygon peg, often though, it wont.

For Linux to work in the round peg market, Linux needs to be a truly round peg.

So what is the big issue?

Linux is not yet ready to be a perfectly round peg. Why? Because their is no efficient large scale method of rolling a set of packages together and calling the result a Linux configured computer. The Linux Standard Base is a step in the right direction: installing libraries and configuration files to the same place across distributions is a good idea, if it ever gets done. But, the LSB does not turn Linux into a truly round peg. The LSB is not enough.

There are literally “Infinity – 1″ different combinations of packages that could be called a Linux configuration (note: I am deliberately not using the word distribution, as that should be reserved for common, predefined, and pre-configured, sets of packages). Until every single one of those combinations can be efficiently specified and created: customers purchasing from resellers will have to pay for a Linux distribution, yes distribution, to be configured on their computer.

We are not going to see this miracle package combiner any time soon. Resellers, like Dell, would be best off:

  • defining a scope of Linux distributions that will configure
  • accepting that may alienate some customers, which they don’t have at the moment anyway
  • make disc images for each supported configuration
  • take each disc image and produce sub-images for each possible hardware configuration that is supported for that image, if a customer wants something else then they should, and probably will, pay extra for it

Or, invest some time and money in creating the miracle package combiner…

,