I like that. One challenge of the PageRank approach is the reliance on authors of HTML documents to determine page quality. It is as if HTML coders (or content authors), who create the hyperlinks that drive “link juice” have been ordained into some arcane brotherhood in which they, and only they, have been granted the privilege of selecting which pages are more relevant than others.

To be fair, the brilliance of PageRank is that it uses meta-data that already exists in the data-set. Hyperlinks are not just native to the web, they are that which defines the web itself. And yet, it fails to capture any wisdom that may be available in the behavioral patterns of other users. As such, it isn’t a particularly user-driven approach (unless you are looking purely at authors as users, in which case, the bottom-up, every link counts approach /is/ user-driven).

As I mentioned when we talked today, I would respectfully challenge the notion that profiling or attributes are going to prove an effective context for refining search. Demographics and psychographics were statistical methods developed by marketers in the mass industrial, mass media, mass consumer era in order to make sense of the overwhelming numbers of customers and audience members accessible through the product and media distribution channels. They were shorthand techniques to try to get a sense of who your customers really were, because the very infrastructure of the era largely prevented direct contact.

In fact, I would argue that for any given product, I’m most likely not like my fellows in my demographic segment. Or perhaps more accurately, it is precisely the ways I’m not like my demographic segment that make me interesting to marketers–because the traditional techniques will reach me just fine for those areas I do fit into the standard buckets. Fortunately, with the net and the age of one-to-one marketing, direct distribution, and even individual behavioral tracking online, companies can now participate and engage with individuals on a one-to-one basis. So, the indicators now for selecting which individuals you want to invest in further (specifically, invest product development and communications dollars for) can move beyond the statistical shorthand of demographic profiling to actual observed behavior on a finely grained basis.

One approach to that is to use Search behavior to drive search results. That is, the fact that Francisco and Tara are “working on a similar project” is far more valuable than the fact that they are colleagues or share particular profile data. So, let’s drive search results based on what people are actually doing, rather than what labels or categories we can apply.

It is an intriguing question if this is a notion that is inherently a part of user-driven search or if its just one of the many ways one can construct a user-driven recommendation engine.

I’ll have to ponder that a bit and comment again in a future blog post.

As I tried to make clear in the conversation at Caltech, I think Salesforce has a lot to offer potential developers: speed-to-market, a rich, complete feature set, and built-in access to a ready market that has proven it will pay for third-party apps integrated with the Salesforce application.

But at its core level of service, it is still not an open platform as I describe. (Which I think we are in agreement about). If users could pull their code from force.com and run it with minimal changes and no loss of functionality at another service provider, then I would agree that it meets the definition of an open-protocol 4th level platform. Those protocols would be the shared APIs of the underlying service layer and the interfaces of the programming languages.

However, my understanding is that a large part of the code written for force.com is in a proprietary language with exclusive force.com APIs. Which means business owners don’t have an easy option to move their service elsewhere.

Of course, this wouldn’t prevent a business from developing its system to work with Salesforce /and/ a number of other services, such as Facebook or NetSuite, but that’s a way to diversify through investing additional resources, rather than mitigating risk by choosing an open platform for the limited resources you do invest in technology development.

What your point highlights is that for a subset of specific functionality at Salesforce.com, the user is locked into a proprietary framework. Of course, they may connect what they build in that framework to other open systems, and can access the application through any standards-compliant browser, but it is still proprietary.

There could be interaction at the level of the application environment (application integration) and also at the level of the user interface (the simple mashup), as suggested by the figure at http://wiki.apexdevnet.com/images/b/be/Level3Openness.gif

There’s no reason why a Level 3 platform can’t offer the rapid time to market of its own high-leverage facilities while still letting developers exploit other facilities, or hedge their bets by reserving some of their code base to run in another environment.