The banders' code was designed for use in banding. It was never intended to be a general-purpose code for North American bird records. In the author's opinion, its use in other sorts of bird records may lead to problems.
Applying the principles in the author's discussion of design goals for bird code systems, the BBL four-letter code is not all that bad a code system. Because it is produced by applying a few rules to the English names, it meets the criteria of being easy to learn and easy to encode. It covers North America and Hawaii, and it is nice and short.
In many of the design goals, though, the BBL code falls short. For example, after reading the list of collisions in the BBL code system, can you honestly say that this system is easy to learn? There are nearly 100 bird names involved in collisions!
Collisions in the BBL system are handled by picking a different, arbitrary code for one or both of the forms. For example, both Lark Bunting and Lazuli Bunting abbreviate to code LABU. Since both birds are common in the West, they are given codes LARB and LAZB respectively.
But suppose a summer intern records a sighting as LABU, and no one notices until the intern has moved on. If we come along later and look at the record, we (or a computer program) may detect that it's not a valid code, but how can we know sure which species was really seen? We might not be able to reach the observer.
People who take wren data have to remember that there is a three-way collision for the code CAWR: Carolina Wren (CARW), Cactus Wren (CACW), and Canyon Wren (CANW). It's probably safe to assume that a Kentucky record for CAWR is a Carolina, but what about a record from New Mexico? In many localities it could be either Cactus or Canyon.
For this reason, the author feels that it is vital to keep the number of exceptions to the rules as small as possible. He once entered a point count survey for the Institute for Bird Populations (which uses the BBL codes) in which over 20% of the records were not legal BBL codes.
When one of the forms involved is rarely encountered in North America, or has limited distribution, the commoner form is encoded as usual and an arbitrary substitute is assigned to the other. This violates the design principle that people who use the code system should not be required to know anything about bird distribution.
To avoid such problems, the author's preference is to disallow use of the collision codes altogether. Then, when someone uses such a code, we will know that they were unaware of the collision. This allows automatic detection of encoding errors by computer programs. Having a computer detect an error takes a lot less time than catching it in proofreading, assuming you have time to proof it at all.
The Bird Banding Lab defines virtually no codes for forms other than species and certain races. Banders have no need for a code for "hawk sp.", because banders will have the bird in hand and be able to key it out to species. But in databases involving sight records, there is a real need for such codes.
Furthermore, a novice observer in Cave Creek Canyon, Arizona, might not know that MOQU is not the code for Montezuma Quail. This is a collision with Mountain Quail, which is more common nationally than Montezuma Quail. Since the BBL does not define codes for gamebirds, who will resolve this conflict?
Finally, please see the list of 684 North American bird names and identifiable forms that have no BBL code.