All articles should be extensively cited with good, reliable sources. Since Wikipedia has no inherent authority, we need to make it as simple as possible for people to verify what we say.
What are good sources? I measure them partly by the ability for the reader to use them to verify the claim we make in the article. Of course this is only one axis of the reliability chart. The other is the legitimacy of the publisher or source, their motivation for being accurate, and the steps they've taken now and in the past to demonstrate trustworthiness. I would tend to reject any source that doesn't pass a minimum level of those tests, and accept any that does (and allow the reader to decide what his or her minimum standards are above and beyond that). For example, blogs are almost always out of the question, even though they are free and easy to get to for everybody, because there is almost no pressure for the blogger to be accurate and reliable. This is pretty straightforward and I'm sure most people understand this already, so I'll move on to the usability.
All else being equal, I categorize citations and their sources in the following ways, in order from most to least desirable.
1. Internet sources that are free to access. Click, and you're there. Maybe you have to register for the site but it is free. They're easy to get to for the reader and easier to verify for fellow editors (and skeptical readers). Examples might be online newspapers and web sites stored by the Internet Archive, or online databases of information such as baseball statistics at Baseball-Reference.com.
2. Offline sources that are easily obtainable. For example, books that are found in most libraries and bookstores, or recent journals that would be found in most university libraries, or newspapers on microfiche. A little bit of a hassle to obtain, but it can be done, and the hassle of going to the local library is better for most than ...
3. Internet sources that are not free to access. Even though you have to pay, you're still only a few clicks away if you're willing to do pay. The barrier of cost is a turn off for most, but they're still attainable by anybody who's interested enough. However if the citation is not legitimate or accurate then you may end up paying a lot of money for nothing. Examples of internet sources that are not free to access include entries out of databases which charge, such as old articles in the online Wall Street Journal.
4. Offline sources that are hard to obtain. This might be an old, out-of-print book that few people own anymore, or maybe a document of which no copies readily exist and you must travel to a particular small town in Texas to view it. These are the most difficult to verify and are most likely to not provide the information that they are said to.
5. Any statement without a source. Banish them all!
Ideally, as much of our citations as possible would fall under category number one. Wikis work when others come by, review, and scrutinize each other's edits. If I say that I pulled a fact out of some obscure book, how are you going to verify that if you can't pick up the same book and see if it actually supports what I said it does? This is to say nothing of the possibility of vandalism by citing nonsense facts out of random books, and everybody sitting back and assuming that the vandal made a legitimate edit (after all, who's going to take the time to read an entire book just to see if a Wikipedia editor was telling the truth?).