March 27, 2003

Scott McNealy

On Tuesday I had the pleasure of attending a brunch, hosted by the Australia-Israel Chamber of Commerce, as a guest of Sun Microsystems Australia. The guest speaker was Scott McNealy, Chairman and CEO of Sun Microsystems Inc.

The topic was about the future of the IT industry which at times was compared to both the phone industry and building your own airplane. Scott was pushing the point that the IT industry has "got it all wrong". The customer is not interested in the parts, they want a solution where the parts are hidden beneath a "blanket". The idea is to outline the solution and assume that the blanket can handle the implementation. Certainly this may be the future but there are a lot of assumptions about both how and when such a vision can occur.

The airplane idea was tied to the development of a systems project where the "owner" buys parts from many suppliers, a wheel here, a seat there, a bit of a wing somewhere else. They then set up their own hanger near a runway and commence assembling this plane. When it all gets too hard, they call in IBM GSA (actually named in the speech) to use some duct tape and string to tie it all together with lots of care but no responsibility. When they finish, the plane is handed over and the owner tries to take off for New Zealand. When it crashes (before takeoff) the system integrator is there with fire-extinguishers, best practice guidelines and a proposal for extending the contract.

Scott also waxed lyrical about the uses of IT when it become ubiquitous and therefore invisible. He mentioned the story about a car negotiating with nearby gas (petrol) stations to get the best price (modulated with the ability to supply milk because the refrigerator reports that it has none) for a given quantity of fuel with no driver intervention until the service station is selected and driving instructions are provided. This segued into a more practical outline of US-wide roaming between Sun offices, the multi-use of office space by having common terminals and smart cards, the extension of roaming to the home and later international spaces.

Personally I liked the independent desktop solution where activating the "dumb terminal" (actually a Java Station in this case - but any thing client host would work) downloads my personal environment. That environment could be Winblows or Linux or whatever because there is a level of independence between the display device and the provision of the environment. We are already doing some of this via Citrix which supplies the "client" side for multiple desktop machines. Scott merely extended this model by making the desktop system no more than a host for the thin client software.

Strangely, given the hosts, there was also some discussion of tagging individuals with unique identifiers. As Scott pointed out, in the US it is normal to use an ID tag on a cat or dog because they are valuable property - surely your kids are even more valuable. There was also a neat example of technology and the car where it reports to "dad" (not "big brother") that the kids are driving too fast and not in the direction of school so that "dad" can talk to the kids and suggest they "slow down and turn right".

The talk hit the right balance of generalism without becoming a hard sell for Sun products. A few times, Scott pointed out that he was pushing standards that were good for the industry even where Sun could not benefit personally from the standard. This came up a number of times in the comparison of .NOT and Java/XML as competing standards. Scott was almost apologetic about the inability of Java/XML to match Microsloth in hosting and supporting viruses.

One of the questions at the end related to the impact of this vision on the industry as a whole. The most obvious impact was on the role of system integrator (no longer required because the stuff integrates itself) which was clearly noted by the speaker who thanked Scott and came from a SI player. Scott also suggested that investing in office space was a bad idea given the more immediate impact of more effective use of existing space and the distribution of the work force - global roaming, he suggested, will completely change the concept of a workplace and for that matter the idea of a CBD. The third impact would be on the component players who could not stretch themselves into solution providers (e.g. Microsloth being "software only").

Overall, the talk was fascinating. In many ways it was a great pity that more of the managerial staff and systems architects were not able to hear Scott expound his personal vision of the future.

Posted by Ozguru at 07:00 AM | Comments (0) | TrackBack

March 18, 2003

MCSE? Why Bother?

Thinking of getting the worlds most useless qualification? No I don't mean a mail order degree from an unaccredited institution. I'm talking about becoming a Microslosh Certifiable Something Else (MCSE). Don't bother. Microslosh are about to put you out of work: Microslosh explores self-managing software.

OK. Here is the deal. Some time ago (1993), some dude got all excited about trying to manage hetrogeneous networks of systems. The result was Cfengine. Sort of worked, quite complex and can be just as unmanageable as the network if not used correctly (i.e. when you get grad students to set it up). This has more recently evolved to v2 with all sorts of bells and whistles (included better control of the grad students). As with all really great stuff which starts out seeming a bit strange, the software is free (as in beer?).

Then the commercial guys started to get excited about this distributed management plus load sharing and application distribution and all that. First cab off the rank that I spotted was N1 (Sun) closely followed by IBM and HP (note all UNIX solutions). Lots of hot air, not much tangible in terms of user stories and real-life test beds yet.

Now all of a sudden, the company that still can't solve the blue screen of death, cannot design secure software, cannot design tasteful interfaces and is the worlds largest convicted monopolist has decided that it will do this too. Now instead of having useless software running on nodes supported by hundreds of MCSEs, we will now have useless software running on nodes controlled by other nodes also running useless software which will be managed by thousands of MSCEs. Oh.

Sorry, false alarm. Lots of MSCEs now required to manage self managing automatically failing software.....

BTW I have worked at a site with ~100 nodes managed using cfengine.

Posted by Ozguru at 12:00 PM | Comments (0) | TrackBack