At the turn of our new century, during the heyday of Napster, the infotech conversation was all about peer-to-peer, or P2P, technology. There were elite gatherings of leading influentials to assess its potential, conferences to bring together ideas and money, new publications to chronicle its development.
In September 2000 Dave Winer wrote a touchstone essay called The P in P2P. “It’s not about P2P, specifically, that people should be getting excited,” Dave wrote. “The killer app of P2P is … People!”
When Groove 1.0 launched I was asked by a media producer to explain why Groove was an important P2P application. Channeling Dave, I said that Groove was important because it empowered people to work together effectively, not because it was a poster child for P2P.
Back then, we lacked the cloud infrastructure to support the loosely-coupled networks of personal services I imagined. But that infrastructure is emerging now. I see it, for example, in the P2P capabilities of Live Mesh.
Nowadays, of course, we rightly no longer regard P2P as a silver bullet. But the infotech tribe is wired to seize on some silver bullet, and so a new one is being loaded into the chamber. It bears an inscription: “the semantic web”.
For good reason! We experience, many times every day, the transformative power of search based on nothing more than the ability to ask and answer the question: “What document contains the word X?” Questions can be a whole lot more interesting than that, and answers can be a whole lot more useful.
On the semantic web, the notion is that queries will traverse ad-hoc collections of data in which the relationships among items of information — within and across datasets — are made explicit. To use the phrase that comes up again and again in these discussions, we’ll be able to “reason over” this web of linked datasets.
We’ll get there, I hope, but let’s not conflate means with ends. During the initial P2P craze we acted as if the P2P web was a goal. Today we see that it’s one of many architectural styles. We use it where appropriate, in concert with other styles, and we’re a bit clearer — I hope — that the killer apps are the ones people make when living and working in a networked world that includes the possibility of P2P.
So it will go with the semantic web, I predict. At the moment it seems like a goal, a grand challenge. And of course it is. But the killer apps will be those that people make when using — and here’s a twist, when creating — the semantic web.
That’s a key distinction. Users of the P2P web don’t, for the most part, build its enabling infrastructure. But users of the semantic web will. Yes, networked computers will weave the basic fabric of linked data. But we’ll need to do the context assembly that makes the semantic web useful to us.
We may be P2P’s killer app, but we’re going to be the kernel of the semantic web’s operating system. That’s why the engineering of that operating system will have to be as much social as technical, and maybe more so.
Jon,
Apropos a lot of what you state in this post, there are 5 fundamental parts to the “Network-OS-Database” hybrid we call the “Web”, currently within scope (imho):
1. People
2. Interfaces (Web Pages)
3. Message Bus (HTTP)
4. Structured Data (Linked Data)
5. Intelligent Agents (Intelligent Processors of Linked Data)
Although we are at step 4, the engagement of “People” is a vital and mandatory activity at each stage.
Step 4. is about the use of HTTP to point to locations/addresses on the Web from which descriptions of things can be transmitted (technically: the act of dereferencing).
Example: http://dbpedia.org/resource/Linked_Data :-) Which transmits the description of the “Linked Data” concept via a familiar Web Page. The only difference is that every link in this page is Hyperdata [1] rather than Hypertext [2] oriented due to the manner in which HTTP has been used to fashion Data Source Names (i.e. like ODBC or JDBC but with record level scope).
“the engagement of “People” is a vital and mandatory activity at each stage.”
Agreed. In general though it’s hard for me to see how the sample queries at dbpdedia or freebase are engaging people to build structured data, or ask and answer questions of importance to them.
The one counter example I’ve found so far is this one from freebase: http://blog.jonudell.net/2007/03/23/semantic-web-as-social-enjoyment/
/That’s/ the level at which these systems need to engage people!
Jon,
I am not in anyway trying to imply that sample queries (DBpedia or Freebase) are the definitive “People Engagement” tools. I am saying that by having “Linked Data”, at this point in the Web’s evolution, we once again have the foundation for “People Engagement” expressable by a newer generation of Linked Data aware solutions.
Here’s my little walk through Web technology and people engatement technologie to date:
1. Interactive / View oriented Web — had the Web Brower
2. Services / API Web — gave us rapid fire generation and sharing of content via Web Serices (blogging, blog comments, tagging, social-networking, electronic discourse etc are included here)
3. Strucutured & Linked Data Web offers us granular access to entities associated with the semi-structured content emanating from stages 1-2 above.
The URI in the href of an anchor tag provides a rich view of the entity “China” across two pathways: Web Page and the Data behind the Web Page. What I am doing for “China” I can do for any Item in my Native or WebDAV File System, Calendar, a Blog Post, a Bookmark, a Tag/ Tag Cloud, threaded discussions etc..
I don’t have to grep-ing for granularity via sed, awk, xpath/xquery, xslt etc., anymore since I have the entities directly at my disposal courtesy of HTTP’s incoroporation into their naming schemes thereby inheriting it’s access and resolution prowess.
Ultimately, I only need to remember my own Name (i.e. my digital data space identifier on the Web) in order to interact with data created by my activities on the Web across a plethora of data generating enclaves (data spaces).
Tagging is another example that will benefit immensely. Imagine if you continued to tag against del.icio.us but enjoyed the benefit of granular access to the meanings of the tags, without changing your del.icio.us interaction patterns.
Of course there are many other examples of how users will benefit from the current inflection without explicit change in Web usage patterns.
“Of course there are many other examples of how users will benefit from the current inflection without explicit change in Web usage patterns.”
I’m sure you’re right. But we need to connect with people where they live.
I met with a local downtown group this morning to show them how their members can collectively manage calendars by publishing and syndicating ICS feeds.
That’s a trivial kind of data linking, but it addresses a problem that everybody feels in their gut. And it relies on nothing more than software they already use and a decade-old standard whose significance nobody has explained.
If we can’t connect on that level — and so far, we have failed spectacularly to do so — we’re not going to get to the next level and beyond.
Jon,
iCalendar is a nice example.
Linked Data enables you to bring iCalendar items to life i.e., you not longer have the burden of expanding the description of a calendar item.
Scenario:
I decide to publish my event attendance calendar, and all I publish are Event Name and Date.
If I publish the iCal data in Linked Data form (i.e. each calendar item has a URI) someone else can expand the Calendar item by adding additional data such as:
Event location, Event Tracks, Speakers, Session Topics etc..
The scenario above can, and will, work in a manner similar to how del.icio.us works today. The key additon is going to take the form of an annotator to compliment tagging.
The annotator will handle the statement/claim formality, leaving users simply enhance items of interested.
What I describe is not going to be unique to Web Applications, it will make it’s way to native OS applications too. Of course, across a plethora of devices etc..
As per usual, I understand your angle re. activation thresholds for time challenged users. But this time around, I do not expect spectacular failure :-)
The underlying substrate is Linked Data, and it is here and finally ready to go
Kingsley