Why not use the Knuth scheme
Where he used increasing accuracy of Pi to generate versions.
First: 3
Next: 3.1
Next: 3.14
Next: 3.141
Next: 3.1415
Next: 3.14159
You get the picture, and I don't need to rack my brain for the next digit(s).
Seemingly borrowing a page from the old, line-numbered BASIC programs of the 1980s, Oracle has adopted a new version numbering strategy for the Java Development Kit (JDK) – one that skips numbers, in case Oracle has to go back and plunk in new code later. Traditionally, Oracle has issued new patches for the JDK on a …
This post has been deleted by its author
So if we have 7u20 and 7u40, and someone finds a bug in 7u20, the fix will be in 7u25? Why not just install 7u40? I don't see the point in an older version having a fix that isn't in a newer version, that way lies upgrade hell, so installing the newest version should always have all the latest fixes, no?
Unless this is because 7u40 isn't expected to be compatible with apps that require 7u20, but that will never happen with Java, I'm sure...
I think that in practice, release 7u40 would be, let's say, less guaranteed to be compatible with software written for 7u20. But I'm not fully informed on this. Another factor is that release 7u40 has an issue date planned well in advance - enterprise customers apparently like that - and, putting the two factors together, there's probably a beta release of 7u40 that you can try out ahead of the actual release, to test whether you're going to have problems with it and your software. Alternatively, you could just do that testing after 7u40 is released and before you install it on your live servers and desktops - but it's likely that 7u40 will be released as a security bug fix release, too, and so you won't want to delay installing it.
Because programs parsing the version number don't expect that. No, I don't know who writes version number readers that crap out on the slightest mishap. Or rather, I know all too well, they are a few offices down the hall.
Sample:
Oracle: Java version "1.7.0_21"
OpenJDK: java version "1.7.0_19"
Apparently, the ".0" currently goes unused...
> No, I don't know who writes version number readers
Unfortunately, I do, me. Not just for java but any software and in SQL embedded in an SQL report to identify machines with old or vulnerable software installed and it is a real pain. You get manufactures changing schemes all the time, even changing their product name. Java is already a pain, and that was started by sun when they started claiming java 1.2 was java 2, some now get installed with 5, 6 or 7 and of course 1.5, 1.6 and 1.7 and a plethora of different product names.