The model of distribution
The validity of a mathematical document A that
refers to a document B can be guaranteed only if B does not change
===> mathematical documents should be immutable
Conjectures can be prooved, theories can be
augmented ===> new versions of a document can be released
Users are not required to maintain forever the
documents they have written ===> URIs vs URLs
(e.g. "cic:/coq/INIT/Datatypes/nat.ind")
The same document may reside on several
servers ===> we need "getters"
that, given an URI and an (ordered) list of servers, return an
instance of the document