Decentralisation

From Change

Decentralisation, or more specifically decentraliased management, refers to the process whereby a system is not controlled and managed from any one central point.

Management is put in place to ensure quality, whether this be of output or behaviour. Management achieves this by providing as a policing and coordination. This works well when the system being managed is relatively small (that is small relative to the resources available for managment). However, as the system grows this problem becomes more difficult and onerous.

Management can refer to explicit administration or control of a system. It also includes the observation of the system as is needed in order to manage a system. But even just observation of a system can become an imense task. The Google search engine has to observe billions of web pages on the Internet and store index results. This task is huge and Google needs to apply many tactics for reducing the number of sites that need to be visited or the quantity of information needed to be stored, yet still maintain high quality search results - this is not always possible and sometimes perfectly valid sites do not feature in the rankings.

Reciprocally, if a decentralised system is left unchecked then chaos ensues. That is, there can be no guarentee that the system is not being exploited by unscroupulous agents or that the system wont simply degenerate. e-mail is an excellent example of a decentralised system. The core system works increadibly well. However, there are no common decentralised management mechanism for vetting e-mail. The result is that the system is heavily exploited by spammers. This is an increasing burden on end users and is being tackled by putting in filters. However, the problem will probably not go away until the e-mail distribution protocol explicity incorporates a vetting mechanism to ensure that e-mails being sent from acceptable sources.

That is, both of these systems would benifit from a well targeted and decentralised management mechanism. In the case of search engines, maybe web sites could manage build their own indexes and track statitics, which could then be verified and aggregated. Thus, the search engine becomes distributed. In the case of e-mail, it seems more appropriate to explicitly maintain a distributed trust system, whereby if a recipient has not granted (in a cryptographically strong way) rights to a send then the sender will either be blocked or incure a high computational penalty.