In this particular case, it means providing an implementation that allows, as closely as possible, everything that the spec allows. For instance, if you look at the code, the Link element is marked as being extensible and as allowing string content, both of which are explicitly allowed by RFC4287 via the use of the undefinedContent element.
Regarding the choice of language, with any new standard, especially one like Atom, it is important to have as many implementations as possible available in as many languages as possible in order to tease out any interoperability issues and to ensure as broad a reach as possible. - James Sylvain Hellegouarch wrote: > >> The goal is a reference implementation. The goal is to be exactly correct. >> Being in a particular language, or even being fast enough to be usable, >> is beside the point. In particular, a reference implementation should >> always choose code readability over speed. > > Fair enough. > >> If the goal is to have a standard, free implementation that everyone uses, >> that is different from a reference implementation and the goals should >> say that. > > I'm sorry to ask but I'm not sure to understand the meaning of having an > implementation which is exactly correct. As we have already seen on this > list, RFC4287 lacks of precision in some context, therefore I wonder what > being "exactly correct" represents. I believe Jigsaw [1] is a an example > of what you mean. Sadly it seems to be hardly referenced anywhere. > > Besides, that does not really answer the question of whay not improving an > existing project over starting a new one from scratch. > > Again, I don't really care about the fact it is built using Java or else > as long as the project ensure bridges to other environment. > > - Sylvain > >
