- Roman commas and italicized commas are distinct creatures and must never be used in place of each other.
- Documents available exclusively on the Internet are introduced using the phrase "at." Traditional documents available both offline and on the Internet are introduced using the phrase "available at." Traditional documents available offline but accessed by the author only through the Internet do not take an introductory phrase.
- Woe betide the poor law student who fails to note that citations to Fifth Circuit cases from 1981 require the month in addition to the year.
I've heard many colleagues, in the face of such seeming madness, say that The Bluebook makes no sense. My personal opinion is a bit different. I think it makes reasonable sense, once you recognize that it's terribly organized. The Bluebook is a textbook example of what happens when you forget to write design documents early on. There are reasonably clear overall principles animating its rules; they've just been obscured by the accumulated cruft of generations of incremental patches.
The Bluebook is, without realizing it, a Chomskyian transformational grammar. At its core, there is a single perfect abstract Platonic citation syntactical form which becomes increasingly specialized as you adapt it to the particular document to which you wish to refer. Thus, given a document, you first write a skeletal citation, and then apply a single deterministic set of transformations to produce a final cite, just as a Chomskyian syntactician would write a deep-structure sentence and then apply a deterministic set of transformations to produce a surface structure sentence.
Now, the problem with Chomskyian deep-structure syntax is that its algorithm is fundamentally irreversible. Given a sentence, Chomskyian ideas of deep structure give you no useful way to parse it. Similarly, if you look at a Bluebook citation, it's not always obvious how it wound up in its final tortured form. There are multiple rules that introduce an "at" before a page number; these rules are all subject to certain exceptions. So if you're looking at a page number and wondering whether that "at" before it belongs there, you can draw almost no useful information from the cite itself. You need to reason backwards, to what the cite would look like without the "at," and then see whether one of these rules (but not the exceptions) would apply.
As best I can tell, this is how The Bluebook used to operate. But somewhere along the way, generations of law students with codifying inclinations forgot about the original rules. Instead, they looked at the places where an "at" might appear and tried to come up with a declarative set of rules for explaining their presence or absence. Since these new rules aren't sensitive to the step-by-step nature of my hypothetical ur-Bluebooking, they're more complicated, messier, and harder to apply than the original version. Instead, they read as a long list of specific recipes to apply in very specific situations. This is the kind of dumb thing you do when you don't have the original design documents to keep you on track.
Worse, when it came time to handle new citaiton forms, successive Bluebook editors have added these new forms by analogical reasoning. Instead of asking why The Bluebook makes a particular choice, they ask when it's mades a particular choice in the past. So rather than respecting the transformational principles at work, they've turned particular quirks of the surface structure into "new" governing principles by propagating them forward.
Which is to say that The Bluebook isn't even a Chomskyian mess. It's a parody of a Chomskyian mess. But it does make sense, in the more specific sense of being explicable. First one finds the general rule. Then one finds the accumulated cruft. As long as one doesn't attribute the details to the general rule, or look for general rules in the details, it's actually sort of reasonable.
Too bad The Bluebook itself doesn't distinguish general rules from specific details.