This section explains how the dynamic web works as a global knowledge base which is selectively distributed across participating sites.

• Because participating sites are free of external controls there can be no viable way of co-ordinating their behavior and we cannot ensure that updates are applied in a consistent way - but we can ensure that all interested sites are notified of all relevant changes.

• For this to work, every change to any individual knowledge element must be represented by a distinct assertion complete with its own provenance. Inclusion of this provenance means that every copy on every participating site includes details such as immediate source and time period of validity. Each site can then use these details to select which of these assertions, i.e. versions, will be included in a page when it is downloaded.

• The directive that causes information selected from the site knowledge store to be included in a page is embedded in the text in a manner similar to that used for links. This directive is a little more complex than a link because it needs to specify criteria for selection of the most appropriate assertion that is currently available at the site concerned. It needs to include criteria for selecting both the property and the subject, i.e. the real-world object or concept descrbed.

• Where the publisher wishes to offer a service that will maintain copies of this information on other sites, this willingness is highlighted on a page together with a link to a dynamic-web subscription form preset for the knowledge element concerned. Any client visiting this page can become a subscriber by following this link to obtain the form. It then adds the URI of the desired subscribing site, and method of payment if required, before submission to the publisher for consideration.

• On completion of any required authentication and payment procedures, generic software at the publishing site will convert the application into a subscription contract. This subscription contract will require the distribution of updates from the publishing site to the nominated subscribing site and nothing more. Each 'update' is represented by an assertion which stands alone as the expression of a new value for a specific knowledge element together with its own full provenance.

• Distribution of individual updates from a publishing site is initiated by a server-side generic process that forms part of the routine management of its site knowledge store. This process responds to the addition of each new assertion by checking to see if it is covered by any current subscription contracts. For every subscription contract thus selected the process will cause a copy to be sent to the nominated subscribing site.

• At the subscriber site all such incoming update messages are checked for validity against current subscription contracts and then, if ok, the new assertion is added to its own site knowledge store.

• Any information held in this site knowledge store can be included in any page forming part of the subscribing site. Such inclusion is signalled in HTML by a call including assertion selection criteria to suit the purpose of the page concerned.

• These are known as collation pages because they can include any number of calls to include the appropriate value for such knowledge elements without following links to the sites at which the updates originate. There is no limit on the number of pages and participating sites through which values of a given knowledge element may be presented to clients.

• From the client viewpoint this dynamic knowledge data is brought up-to-date from this local site knowledge store on every visit/refresh. This differs from dynamic pages based on a conventional site data base in that there is no reliance on a dedicated server-side application system. Locally sourced assertions sit in the local site knowledge store alongside those sourced from any number of publishing sites. It is up to the designer of the subscribing site to decide whether the end user is advised of the source(s) and/or update policy for such information and, if so, how such provenance details are presented on any given page.

• All these knowledge elements originate as locally sourced data through a dynamic web knowledge entry form filled in by some user of some site. What makes the dynamic web a coherent global entity is that the entry format is standardized to ensure that each unit of entry is treated as an "indivisible and immutable version which is semantically complete and has its own provenance".

• Adherence to this principle enables contract-based distribution across the web to be handled entirely by generic software. This software has two components each of which have two roles as shown in the following table:

Knowledge distributed through the dynamic web goes through the following four phases:

Specification. This is where a publisher adds the definition of a 'type of knowledge', i.e. property, to the global ontology.

Upload. This is where a client user enters information in the form of assertions for uploading onto the dynamic web.

Storage. This is where information in the form of knowledge grains is uploaded to a participating site and held for use in subsequent page downloads.

Download. This is where information forming part of the dynamic web and held as assertions at a participating site is inserted into a page immediately prior to its being downloaded to a client.

Control of the service through which information is distributed across the dynamic web comprises the following four phases:

Offer. This is where a publisher uses a website to inform visitors of the dynamic knowledge data distribution service on offer.

Agreement. This is where a subscriber specifies an interest and enters into an agreement to receive updates at a nominated subscribing site.

Delivery. This is where information in the form of assertions uploaded to the publishing site is distributed to other participating sites in accordance with subscription contracts.

Administration. This is where accounts are maintained for services performed under subscription contracts including statistical and other reporting obligations as well as billing and payment where appropriate.