...
2) have a fixed length, and limited syntax. Why? If we pick a type of identifier with a large identifier space, but an extremely rigid syntax, the identifier is then more easily found in html pages, and manipulated by such things as greasemonkey, or libX. In addition, if we are dealing with an identifier that has been printed, and then needs to be recognized by OCR (think about a cell phone camera grabbing the doi's from a printed pdf, and bringing them up on your laptop screen), the job is much easier (I mean, actually possible) if the identifiers implement a really very rigid syntax. arxiv identifers, namely, arxiv:xxxx , are really quite good in this respect, I think.
...