[
Lists Home |
Date Index |
Thread Index
]
It is interesting to compare Fielding's dissertation
with its claim for the novelty of REST with the earlier
work on Dexter. Identity as intrinsic property
(the global namespace assertion) vs idenity as a derivable
property based on function comes up often.
"In addition to the data model, Dexter also specifies a set of operations on this data model. The resolver function is responsible for resolving a component's unique ID given a component specification. This indirect addressing is used to cope with dynamic changes of component IDs, i.e., when components are edited. In the simplest case the component specification is the component ID, in which case the resolver function is the identity function. The component's ID is then fed to the component accessor function, which is responsible for mapping that ID into the data object assigned that ID. Furthermore, there are functions for determining the interconnectivity of the network by LinksTo and LinksToAnchor operations. The first resolves all links to a given component whereas the latter resolves all links referring to a given anchor within a component. "
http://www.dbai.tuwien.ac.at/staff/herzog/thesis/node25.html
Whether one model is more complete or just the another model in different
terms and explanations would be an interesting thesis for some one.
len
-----Original Message-----
From: Mark Baker [mailto:distobj@acm.org]
Anything-goes RPC is basically the null style because it defines no
constraints on component interaction. I assume that's why it didn't
get its own subsection in section 3. OORPC at least defines the
constraint that components are objects with identity.
|