I agree that what I have described from a user experience perspective is able to implemented by a good programmer and using information feeds available today.

 The critical difference is the the API supporting these applications are predetermined and conform to expected schemas.  But what happen when you take your device to China, or Africa, or Australia.

The composer needs to be able to search for most appropriate services that suit your needs, putting together a unique application instance for you and your context.  For this to work, the standards and practices of the Semantic Web need to become ubiquitous, not just implemented in pockets as the early semantic efforts are likely to be.

However, it is good to see more applications like the example provided trying to understand customer intention first.

with regards to social networking functionality, I think definitely, with the most obvious example just being able to highlight my friends favorties/preferences and the more advanced being able to use directed graphs based on social networks to optimise the experience on multiple levels.  But that's a big topic for another time.

What Ilya describes is available today. Go to http://travelocity.roadtripwizard.com.

This type of solution can be closely integrated with social networking functionality as well. Feel free to let me know what you think.

Yea that semantic web stuff is a big deal.

I think one key point overlooked in the semantic space is the importance of the URI. More specifically, how much data we retain in the URI. This quality is what I have been calling Isomorphic URIs -- and how isomorphic a URI is depends on its ability to describe what is being returned in html/xml/etc.

Simple changes to how sites treat URIs and we could begin to actually use websites as contextual DBs with generic interfaces. A lot of this has to do with an underlying search mechanism as well...

Snip:: 

Resource Request and Return Relationships
A resource request (REQ) as a URI and a resource return (RET) as data will express an isomorphic exchange where elements of a URI can be correspondingly mapped unto the returned data structure and where a data’s structure can be correspondingly mapped onto the elements of a URI – simply, the process must be reversible. Many transformations between the original REQ and RET may occur, though a corresponding mapping must largely be retained to qualify as an isomorphic exchange.

Today, many such corresponding mappings of URIs and data are implemented on the internet — however, the strength of isomorphism and the methods in which the data is mapped varies wildly. The varying degrees of isomorphism and data mapping, though usually URI-compliant, make it difficult to reliably request data from a given resource without intimate knowledge of how that resources is structured…

If interested, feel free to read more about Search-based Contexual Routing via Isomorphic URIs @ futureprogress.net. 

>||;)

Well, your answer is turning me into a believer. So, I am learning more about Web 3.0 starting from this iconic Web 2.0 entity. Thank you for the enlightment.

I think the challenge you are raising is the challenge of interfaces to these composer type of containers.

I agree with you that I dont always feel like 1 or 2 types of food.  However, I see it as a simlple matter of being able to put your query as Find Springrolls.

A proper semantic web solution to this can resolve springrolls to being food with an attribute or link of Chinese and pass that information across to the appropriate directory service in place of your preferences. 

This does also raise some interesting questions about accepted "best-practice" for inflating the meaning of a request.  If someone provides additional information such as "Spring Rolls" does the composer;

1. Use the Style of Chinese in place of preferences; 
2. Use the Style of Chinese + preferences; or
3. Just use the preferences.

I think the industry will be consumer driven in this regard and rapidly align with some common sense rules. 

Ultimately, this should enable more choice to the user, with the user able to have preferences if they want, or not.  If you feel like something specific, assert it with your intention.  If your open to new experiences, you can assert that in your intent too.

I can envisage a little feature on these things called "Surprise me" which is like an override of your preferences/tastes to see what happens

James,

 An interesting question.  First off, I should qualify my answer with the fact that I am an optimist.  So people may want to take my responses with a pinch of salt.

 With regards to when we will see these "applications", I think that we need to consider a new concept.  I think we are more likely to see the rise of composition containers which can generate unique, once-off experiecnes based upon a number of factors - presence, network services, device capabiliites, etc. And I think these are a long way off.

 However, we are seeing a maturity of tooling around web 2.0 enabling mash-ups, etc that provide ever increasing sophistication in "guessing" what bits of data you are trying to associate.  What is missing is a set of standards for representing meaning that platform builders can rely on.

It is going to be the standards which can represent meaning and associate intent to methods which will be critical.   Much like the basic syntax of XML has permitted the web 2.0 exploistion, the consistent representation of semnatics will enable the leap onto the Semantic Web.

But back to the question - how far away is that?   There is certainly pockets of research into these standards in academic institutions and product companies which are playing around with possibilities and some of that effort is being organised by the W3C.  RDF, OWL and XML seem to be the emerging leaders for the syntax standards for Semantic Web and is stablising as the basis for SW.  To support the academics and product R&D open-source and commercial product tools are emerging. 

For the scenario, the semantic standards will have to be bedded down and penetrate into web service standards, becoming a ubiquitous way of expressing data.  This is probably on the 5+ year horizon given how fast dominant companies tend to move in this space.  Maybe some new start ups will emerge to take ownership of the Find verb away from Google based on their research but I would be surprised if it could be commercialised on a large scale any time soon.

Companies doing this now? I would say the leading adopters of this work would be the leading commercial search platforms (not suprisingly) with companies like Endeca and FastSearch doing lots in this space.

I am very interested to find out from the online community if anyone is starting applying these concepts to run-time composition

 cheers

 Ilya

This brings up another important point. To get better contexual understanding of what the user may want, the computer is going to have to track and analyze more user data than ever before. Right now, people are throwing a hissy fit whenever any data is tracked, but I believe that it will become the norm to track massive amounts of user data in the future.

Perhaps the tools and user information database needs to be a local resource instead of a web resource. Then Google (or any other company) need not know what you are up to but a tool they develop can still have knowledge of you and your patterns (in order to predict your requests before you make them) and/or offer other helpful contextual information. This local "database" would have to be pretty secure though, as it would be a MASSIVE target for hackers and other malware authors.

Ilya,

the concept of having your personality stored on computer/network is the basis of this Web 3.0 thingy. (It is almost the same with Web 2.0, I know...) While everybody seems to be thrilled with that fact, how it makes your life easier, how machines can take intelligent guesses about which restaurant you want to eat in, etc...; few people deal with the fact that "making your life easier" equals to "incapacitate you of making decisions" in some cases.

That might seem a little harsh, but for someone who has a strong idea about the place he wants to eat in (e.g., he would like spring rolls NOW, but he usually eats Italian and Mexican food, so his profile contains those), the Web3 version of picking a restaurant is surely irritating.

My point is that you should have the choice of using your mind and some help from the computer/Web or only the computer/Web. This is like having the choice of skipping an online ad, when you checks your favorite new portal, or the choice of not using the Flash version of a site if you are interested in content not design. (OK, these examples might be weak, but you may get my point).

So if you prefer it Web1-style, you shoud be able to do so. The choice should be yours and yours only.