Distributed Publication
Submitted by JeniT on Fri, 15/10/2010 - 15:47 | Updated on Mon, 28/03/2011 - 15:56***Status:** Working Draft*
*This document forms part of the data.gov.uk guidance about publishing Linked Data within the UK public sector.*
One of the unique selling points of linked data is that it supports distributed publication, whereby people who own the data about something can publish that data with only minimal and incremental coordination with others who are publishing data about the same thing.
The following figure illustrates the distributed publication of data about schools, using the linked data approach.

## URI Sets ##
In the centre of the diagram is the URI set for schools. **A URI set is a collection of [URIs](/resources/uris) for particular kinds of real-world (or abstract) things that follow the same pattern and are usually based on some unique identifier for those things.** For example, the URI set for schools follows the pattern:
http://education.data.gov.uk/id/school/{unique-reference-number}
where the *unique-reference-number* is the number assigned to each school.
URI sets themselves also have URIs; in this case, the URI set for schools has the URI:
http://education.data.gov.uk/id/school
## Datasets ##
On the left and right of the diagram are datasets that are created by different organisations: Edubase on the left and Ofsted on the right. These datasets can vary hugely in size, content, timeliness and so on. In this example, a single Edubase dataset contains information about multiple schools, whereas each Ofsted dataset contains information about a single school. Not shown in this diagram is the fact that each dataset may contain information about things from multiple URI sets: administrative areas as well as schools, for example.
When a dataset contains information about a real-world or abstract thing that has a URI in a known URI set, the dataset should use that URI to identify it. This enables third parties to easily pull together the data from different datasets.
### Canonical Datasets ###
There is one set of datasets that has a special role. When someone requests data about a particular real-world or abstract thing using its URI, the web server responds with a pointer to a document that contains data about that thing. This is shown in the bottom part of the diagram: the URI for a school points to a document that contains data about that school. The owner of the URI set gets to determine what these canonical datasets contain.
## Governance ##
Anyone can define a URI set. Useful URI sets are those that are supported, such that:
* the URIs are persistent over a long period of time
* requesting a URI within the URI set results in a redirection to a document that contains useful data about the identified thing
* requesting the URI for the URI set itself provides documentation about the scope of the URI set, guarantees about its persistence, and the way in which identifiers are assigned to things
It is up to the owner of a URI set to determine where a redirection from the URI for a real-world or abstract thing points. It may be within the same domain as the URI set, or a different site. It is not useful to invent URI sets in web space that you do not own, since in that case you cannot control what redirections or documentation is available at those URIs.
Because anyone can define a URI set, a given real-world or abstract thing may have multiple URIs. For example, there may be multiple URI sets for schools, perhaps using different numbering schemes. There are two main reasons for this:
1. Different organisations may adopt this approach at different times; the first may need to define URI sets to support the data that they produce, while the second is the true owner of the identification scheme (the organisation that assigns numbers to entities, for example). Once the true owner of an identification scheme begins to support linked data, any secondary URI sets may be deprecated, with the URIs simply redirecting to the new URI set.
2. Different organisations may in fact be identifying different aspects of a real-world thing; for example, schools might be identified as buildings, as organisations, or in terms of the level of education that they provide. In this case, the distinctions between the URI sets are important. It's possible for a given school organisation to occupy several buildings and to provide education at multiple levels. It's also possible for two schools to occupy the same building. The use of different URI sets enables data providers to be specific about what kind of school they mean.
Anyone can say anything about the things within a URI set within the datasets they publish. It is not necessary to own a URI set in order to refer to items within it. Where there are multiple URIs for a given thing, the owner of a dataset can choose which to use, and may use several.
Providing useful data at the end of a URI makes it a more valuable URI set, and therefore will encourage dataset publishers to use it. Using common URIs makes datasets easier to reuse and combine with other datasets. These facts combine to eventually encourage convergence on common, useful, URI sets, without the need to decide what they all look like up front.
## Services ##
The second figure, below, demonstrates the way in which services can be built on top of these datasets. An important principal for open government data is that anyone can build services using the datasets that are made available in this way.

Service providers add value to datasets by, for example:
* providing data in different formats or through interactive visualisations
* analysing and summarising data to provide aggregate information
* combining them with other datasets
* augmenting them, perhaps by converting values or performing reasoning over them
* providing extra means of accessing the data, such as SPARQL endpoints or APIs
* collating commentaries and annotations from social networking sites
Some publishers of datasets may provide these services as well, but the two roles should generally remain decoupled as this helps to ensure that datasets are in a form that can be reused more generally.
The results of a service are usually new datasets, which may themselves form the input to other services. For example, a service provider might pull together basic information about nurseries from Edubase with Ofsted reports, and enable the public to form communities around them, providing comments and reviews on the nurseries. A second, hyperlocal, service provider could then select just the nursery information for a particular area for use on their own website.
## Other Resources ##
* [Designing URI Sets for the UK Public Sector](http://www.cabinetoffice.gov.uk/media/301253/puiblic_sector_uri.pdf)