This is something that has bothered me for ages. I wasn't sure it could be made up as a question here, until Olin Lathrop suggested it to me in a comment to another thread.
We all know what we are talking about when someone mentions, for example, an 1N4007 or 1N4148 diode. The first is a general purpose rectifier with 1A max forward current and 1000V max reverse voltage, just to be explicit. And this information is completely independent from the manufacturer of the specimen we have at hand (I'll bet anyone with a bit of experience in EE would go crazy if he saw, say, a DIAC labeled as 1N4007 produced by a reputable manufacturer).
So we know there are "standard" part numbers which corresponds to devices with well-known characteristics. As Olin pointed out in that answer to the question in the thread I mentioned above, although we all know what a standard part is, sometimes different manufacturers give slightly different specs for the same part. Still, an 1N4007 is an 1N4007, this little differences notwithstanding!
So my question is, what defines what a standard part actually is? Historical reasons (the first manufacturer got the part number; second sourcing the part to others made that part a de-facto standard)? Industrial agreements (some manufacturers decided it would be good not to do cutthroat competition on low-tech parts, so they decided to standardize)? Official standardization organizations (Olin mentioned JEDEC in that comment; may it be some other institution as well)?
To really gild the lily the answer should also provide references to possible official documentation about common parts or procedures (if they exist) to standardize a part.