Amit Mital kicked off TechEd Australia 2008 today with a keynote presentation on Microsoft’s view of how software and services will develop in the future, particularly in relation to their new Live Mesh offering. There is a good summary of his presentation on the TechEd New Zealand site, it seems they got an identical opening keynote. For someone who loves networks he sure doesn’t seem to like professional networks!

There was one flow of logic which struck me in his speech. Moore’s law is still holding true, and computer hardware is continuing to double in processing power every 18 months. This computer power is also appearing in more and more locations. But when was the last time your network doubled in speed? What about doubling in speed to each additional node? This rapid processing power increase has meant two things that are obvious even today:

1. Computers are islands of computing power - There is no seamless transfer of data between your devices. You work on a file at work, email it home, download it at home, work on it and send it back.
2. Deploying local machines is too hard - Each branch office needs a rack, servers, backup, redundency, configuration, support, licencing…

Microsoft’s solution at a high level is the Mesh stack, the structure of which can be seen in the slide shown here. The fundamentals are that local software is fast, hosted services are convenient, so lets tie them together with an API and we get the best of both worlds. The trick is getting the balance right, where does a local application end and the service begin? How do you split the business logic? How do you provide offline access and quick sign-on to new devices? Hmmm…

Microsoft’s current practical solution is to re-write most of its server packages to allow hosted delivery. Hosted Exchange is an obvious flagship for this. Google have taken a different approach. They believe that all you should need on your desktop is Chrome, essentially an all-purpose thin client rather than a thick client on a drip feed.

So who is right? Well I am betting things will converge on a middle of the road approach. Implmenting with current technology I would say that javascript, a web browser and some sort of XML interface would be the best way to go. A few things need to develop from here:

1. API’s need to be standardised and built into the browser (or OS as these merge). Something like Javascript libraries, but compiled, lightening fast and highly reusable. Chrome is getting there.
2. Data transfer needs to be better than XML. Think highly compressed, encrypted on the fly, but quickly decoded into a human readable format if necessary. Microsoft’s MeshFX is getting there because it has authentication and other services built in, but it needs to be open like SOAP.

So I guess the race is on! Google will take Javascript to it’s limits, Microsoft will try to blow us away with it’s feature set. When will they sit down and standardise on the next generation of javascript and data format?

SAAS is a cheap delivery model right? Well it depends if you are a technical or a business oriented person. There was a very interesting article on Business Spectator today regarding the costs of deploying SAAS software.

One of the most interesting statistics for me was that Salesforce.com (one of the few genuinely profitable SAAS success stories) spends half its revenue on sales and marketing. When I think about it it makes sense, but I guess coming from a technology background it isn’t something that I would have initially thought of. The scary thought that comes after this is what percentage of revenue is taken up once you include support as well? The even scarier one is are you really achieving scalability if these represent the vast majority your costs?

As the article points out, there are three economic fundamentals:

1. How much it costs to attract new customers (known as subscriber acquisition costs)
2. How much money can be extracted from those customers in regular subscriptions (known as “average revenue per user”, or ARPU)
3. How often subscribers drop out and have to be replaced (the churn rate)

For point 1 I think SAAS companies still rely on their old mantra that the SAAS model generally is still struggling for widespread momentum. I think this still holds true, but perhaps these days more because of the costs of migration and training rather than lack of market acceptance or knowledge. The sales and marketing costs are therefore quite high as they are pulling lagging and inflexible customers into what is now a competitive market place with low barriers of entry.

These low barriers of entry tie into the last two points. Companies can theoretically treat SAAS software like they would their telephone company and switch at any stage. This means that new entrants can come in and undercut the established players, keeping subscription revenues down across the board.

The interesting part for me is that online classifieds is somewhat like the SAAS model. It relies on subscriptions and has a low barrier to entry. Why are they so profitable then? I have a few ideas:

• A new twist on the established classifieds model, rather than a whole new model
• Industry maturity leading to a consolidation of players (and profits)
• The difficulties in targeting consumers online

To be honest I really feel like I am missing something here. Can anyone help?

Let me say this from the start, I think Open Source software is the way of the future. Let me also say that I am not a lawyer. On that point I find it incredibly ironic that the urban definition of the IANAL acronym disclaimer directly references the GPL. This is even taken one step further with another more ridiculous acronym created during a GPL discussion, which conclusively proves that giving acronym loving nerds a sniff of legal jargon is a recipe for disaster.

It also helps explain my point, the GPL isn’t the Open Source saviour some people think it is. I hate to list points, because the people feel they need to find a way to argue against each one rather than the logic as a whole, but I am going to do it anyway:

1. Money = Evil - Any efforts to profit from Open Source development work is treated with scorn. Comments such as “To release a non-free program is always ethically tainted” frustrate me. Equally promoting an Open Source product does not give you licence to trash commercial software. They are your competitors, if you think you are on their level then treat them with respect.
2. Discourages Integration - The GPL is brilliant for making utilities. Compilers, databases and graphics programs are all essentially utilities that you interface with in a certain way, but never extend or deeply customise for your own purposes. Deep integration is one of the biggest competitive advantages that Open Source has over commercial software packages, so why make it hard? For example SugarCRM allows web service integration; but even modules, templates and dashlets that integrate within the existing API’s are considered to be covered under the GPL. Is this really a deep extension of the core product?
3. Patent Protection - Patents are either loved or hated, usually depending on whether your name is on one or not. Regardless, the fact is that they are not going away. Open Source products are just as vulnerable to patent infringements and litigation as commercial software is. As the lines between Open Source and commercial work continue to blur, it is emerging that corporate indemnification is almost becoming a quality assurance stamp. A community cannot offer indemnification, so they really need to focus on their competitive advantages. Stay away from heavy duty licences that just muddy the waters for smaller businesses and institutions, don’t forget a hatred of licences seeded your whole industry!

As with anything legal there is no ironclad solution. As far as I can see the solution is to make it as easy as possible for people to contribute maximum value with minimum overheads and receive value for whatever purpose they desire. Rely on the fact that producing a commercial product that is 99% Open Source is not a safe, competitor-free business model!

This was a project that I completed for the 1st year Advanced Software (SOFT1901) stream at Sydney University in 2002. It is a good example of some of my java capabilites, especially applet work. I will post the full applet online in the near future. Click for the full-size image to see more detail: