Back in those exciting days when evangelizing Linux and open source was still exciting, I almost got lynched for a simple statement at a Linux conference. The statement: “Linux will only be important when no one cares“.

Hindsight being what it is, I’d like to think I was right, at least on servers, in embedded devices. Let’s hold judgement on desktops for a moment.

Linux is now just expected. Kind of like VMS in the 80’s, Windows in the 90’s, if you aren’t doing Linux, you are ignoring a significant and increasing part of the market.

So what, you say? (And you’d be right… see, no one cares).

Well the issue facing Linux is a new and increasing confusion on what Linux is. In the last year as a consultant, I’ve met with 5 companies who wanted to know: 1. which Linux to develop for? and 2. how to get out of the loop on maintaining multiple, development-incompatible Linux flavors.

Well, there’s no good answer. We still have the Linux community (the last remaining “cares” group ready to extol the glories of their favorite distribution, be it Red Hat, Novell, Ubuntu, or my choice for a new distribution “Britney Spears Linuxthe distro with nothing hidden. It’s not trivial for companies, especially small companies, to pick and create support for all of the possible choices, nor does the current “standards” cover all the possible contingencies. (But at least the FSG is trying). BTW, big companies face the exact same challenge; they may have more resources to throw at the problem.

Already, the picket signs are up. “Just make it open source and we’ll do the rest”. Yeah right…

Believe it or not, it costs real money to release a product to open source. It can cost real money to contain potential damage from bad implementations. or instance, say I release a device driver for a new storage device. I open source the Red Hat version of the driver, and it gets ported off to my BSpears Linux. Some unnamed company decides to use the driver and my hardware device to store their customer database, in spite of the “not supported” comments. When the device hits 80% full, the ported driver has a seizure and crashes, taking all the data with it. Guess who gets blamed… it ain’t Britney.

So, what’s the right play here? Should I go with the Red Hat dominant market share, the Novell (a newly-indentured servant of Microsoft), some version that no one in my market uses? Should I focus on Germany (one answer), Japan (different answer) or the US?

A long while back (in my SGI days) I came up with a handful of questions we asked groups when they wanted to release somethng into open source. I should dig them up and run them as a blog sometime.

So, what’s the hindsight going to be in 2010?

as always…


Never confuse data with storage.

It’s so obvious when you say it. It’s also amazing how often the two get confused, especially at the C-levels in companies. Intertwined, intermingled, unseparable.

Storage is basically the ones and zeros. Storage is where the btis live, not what they mean. Data, one the other hand is what the bits mean, not where they live. Consider the old world of Oracle and raw storage. Only when it was processed by Oracle did it become data.

Thus we can have tiered storage, based on tech specs, in places like Wide Area File Systems, replicated storage. Thus we can have content aware data, e-discovery.

Today, certain corporations are trying to blur this distinct, and thus limit the market choices. Often large storage companies are making the claims that they deliver data… yet sell storage. Be very careful in going there. Often the exciting work in data (as data) are found in small start ups, well at least before they get acquired by big companies.

01000100 01100001 01110100 01100001 00100000 01100001 01101110 01100100 00100000 01010011 01110100 01101111 01110010 01100001 01100111 01100101 00101100 00100000 01110100 01101111 01100111 01100101 01110100 01101000 01100101 01110010 00100000 01100001 01101110 01100100 00100000 01110011 01100101 01110000 01100001 01110010 01100001 01110100 01100101

as always, comment away