Wikipedia, other encyclopedias, dictionaries, etc. as we know it, have the following in common: they are a list of usually single words (entry words, items, index words, etc.) associated with longer texts (articles) for which connection (whether explicit or implicit) the whole collection is sought. The way the number of items in the list is growing is either by inclusion of a new item that an author knows enough, or by the inclusion of a new word as a stub that a curious reader is not knowledgeable enough and wants to know more of. Each entry word first is found somewhere out there (in the context) and is then decontextualised, lemmatised, etc.) to be included in that list. If there are homonyms, the original context of each occurence is partially reconstructed and marked by disambiguation.
Context, however, should be considered to introduce additional relevant points (sometimes the originator for example), and if you go on without exactly specifying the relations that a given context reflects, you should have a lot more contexts shown in a structured fasion than normally found in dictionaries and/or in encyclopedias (compare: categories, senses, semantic web, etc.).
Given that wikipedia and its sister projects are word-centered devices you can hardly go beyond basic grammar terms and the alphabet used a) to bring an entry word to a common form (noun, single word), and b) to sort them according to a single („senseless”) criterion that keeps resulting in an alphabetic index only to rely on in searching.
2. The text/article
Since on learning about the world and/or words all of us progress from known items toward unknown items by establishing various new connections and sorting the input individually, it may be desirable to include a ring of pointers that a) take you from scratch (trivia, dictionary entry) to the latest advances connected to an entry word, b) allow you to check for covering the complete process of acquiring knowledge in a given subject in succession. And context therefore should not be dispersed aroud the free-floating text. In other words, knowledge associated with an item should be graded and presented accordingly. Using a hypertext structure is fine, but not using other text processing tools such as concordance (KWIC/KWOC) programs to recreate and/or remodel textual information is not quite comprehensible with so many computers at hand.