Sunday, February 29, 2004


The web is an incredible rich environment for information sources, however as a medium for a rich client experience it is woefully lacking. The current set of technologies fall into two extremes, HTML (fast, easy to program, and runs everywhere) and Java (complex to download, complex to integrate into the web, and complex to configure to run everywhere).

The rate of information growth is approaching exponential and the information about information is growing, but search engines still deliver information as single dimension lists!. The Semantic Web will be stillborn if we do not break free of HTML and deliver a common rich client environment.

The need for a rich client environment that, runs everywhere, is fast, and easy to program is holding back the next wave of innovation and more importantly delivering understanding to the users of information. There are a great many solutions out there but they are all falling short between ubiquity, easy to use and fast. This is the next big challenge for the web and the Semantic Web is DOA without it - who is up to the challenge?

Thursday, February 19, 2004

Linkedin - activity picking up..

I started with Linkedin as an experiment and I have actually seen some benefit from it. I have forced myself to use Linkedin even when I could have used other avenues just to see the effect and to see if it actually kick starts something.

The benefits have not been enough to see any dollar value associated with it as the links I could have made other ways. There may be a critical mass effect at work here, once the network gets to a certain size and you have a certain number of contacts the value actually kicks in. I would be interested in hearing other users experiences.

Tuesday, February 10, 2004

Off to Emerging Technology...

I am off to ETech for the rest of the week, I am speaking on Wednesday at 5:15 "Services not Software or why are you still moving mountains". Fun topic hope to have a lively discussion with everyone.

Monday, February 09, 2004

Shared infrastructure why?

This is one of the questions I get asked, and one of the easiest ways to explain it has been to use a simple use case.

Currently there are two attachment formats in use for web services, both supported by large companies, MIME and DIME. This caused issues for several of our customers as they had customers that supported both. They either had to a) Get everyone to support the same format, b) Support both formats for all the services they built, c) Tell their customers that this is the format we support and hope their customers could support both. Neither of these solutions is either particularly economic or customer friendly.

To solve the problem Grand Central implemented a configuration option that allowed users to specify the format they wanted their attachments delivered in. This code was extensively optimized to ensure that the actual impact on performance was only a few millisecs. The total investment in design, implementation,QA,and installation was about 4 man months. Now every customer can use the feature and is freed from having to worry about what attachment format everyone else is using, and if they change their format they also do not worry.

The engineering behind this is fairly simple and many will cry I could do that. The answer is yes you could do it but why - what is the value to an organization to build this and similar features. If 50 organizations need to build this the cost is 200 man months, at $10,000/man month we are talking about $2,000,000.00. This is a simple example more complex features cause the numbers to increase dramatically.

While not everything should be in the shared infrastructure this simple example clearly shows the value of shared infrastructure in removing direct costs from the enterprise and reducing friction between organizations.

Tuesday, February 03, 2004

Social networks a peer to peer problem?

There have been several (many) posts apophenia: why Orkut makes people insecure bashing Orkut for many reasons. The majority of the posts are valid but I do not see that the problem is unique to Orkut.

Trying to create a general set of semantic information to bring together large numbers of people is inherently an almost unsolvable problem the way Orkut and most of the other networks addresses it. The meaning I assign to any of the fields I select in Orkut is probably different from how others assign meaning. This is a classic problem with using shared semantics - i.e. how do we arrive at the shared semantics. Having a centralized authority specify the semantics (and in the case of Orkut a limited set) is the wrong approach (IMHO). What needs to happen is that the infrastructure should enable peer groups to form around common semantics (or Topic Maps).

There needs to be shared infrastucture to enable social networks but the actual semantics need to be created in a peer to peer interaction between individuals. Is something like Easy News Topics a good start? perhaps, but we need to figure out how to make this easy to use, while delivering benefits to users.