said by new2fios:The most I have ever seen is 4 (Microsoft)
...because that's the limit imposed by their operating and file systems.
said by new2fios:but usually code released to the public has a 2 digit revision code.
I'm looking at the Sun Java runtime—probably one of the most deployed pieces of software in existence—and it has six levels. I regularly see open source Linux projects with five or six levels (because they don't have the MS constraint of four).
I'd submit that classifying all changes into one of only two levels of significance is not granular enough. The whole point of having multiple levels is so that one can judge the relative significance of changes between two versions; the more "bins" of significance, the better that difference can be judged.
Look, I'm not denying that, the more levels, the longer it takes to "decode" and the more difficult it is to compare two versions at a glance, and that the use of nine levels
is highly unusual, but I think it is a bit harsh to call their practice absurd just because it is over some arbitrary limit that you've decided is enough.
What perhaps
could be considered absurd is not publishing the criteria used to classify changes. I've never found such a source for Actiontec/Verizon. If there no context of what constitutes, say, a fourth versus a seventh-level change, the whole point of classifying is rendered useless, and you might as well just use one level. For example, as an embedded systems dev myself, I publicly document the fact that fifth- and sixth-level changes are those that have no operational significance (HMI, documentation, etc.), so a customer can instantly see that a change in one of those levels need not be retested.