On Feb 15, 2010, at 1:13 AM, Ralph Goers wrote: > I'm not on the legal affairs committee, the board or the logging PMC so my > opinion here is just that. > > I haven't looked recently at how prudent mode is implemented (although my > experience with file system locking tells me that no matter how good it is it > will fail on some systems) or what was done to implement it in log4j. It > doesn't really matter. As I understand it Ceki is still on the logging PMC. > Although he didn't explicitly state it, I take his statement below as a veto. > http://www.apache.org/foundation/voting.html clearly indicates that the code > needs to be removed until the veto is resolved. > > I would also suggest that Ceki needs to be careful in not vetoing things > simply because they are already in Logback as there is a potential conflict > of interest here. If Curt's account is indeed true then I'm not convinced he > did anything wrong, although a review of the code might persuade me. There is > nothing necessarily wrong with looking at some code, learning something from > it, and then writing code independently, but you can't just take the code and > modify it a bit. The difficult part with this is that if the code in question > is small and there are a limited number of ways of doing something then > everything might look like a copyright violation even if it isn't. Wikipedia, > at http://en.wikipedia.org/wiki/Copyright_infringement, provides a good > discussion of the issue. > > Again though, given Apache's rules on vetoing I don't think all this even > matters in this case. Somehow the code needs to be reverted and then > reimplemented. Curt's analysis in the Jira issue, and even providing more > detail about what nio methods to use, would easily be sufficient to insure > that there is no copyright problem. > > I am also concerned by Ceki's observation that there are insufficient tests > in log4j to test the added feature, not from a copyright point of view but as > a general rule. How do you know the code in question even works? If I was on > the PMC I would veto a code commit of a new feature that had no tests. > > > Ralph >
The first comment was an question which I answered . After Ceki disputed my account, the code was reverted quickly. As for tests, there were tests for isPrudent, setPrudent and that setting prudent mode and doing some logging didn't explode. There were not tests to confirm that prudent mode does what is claimed, as noted in both the bug report and javadocs for setPrudent, the effectiveness depends on implementation specific behavior. The Javadocs in the patch clearly discourages users from having multiple JVM's write to the same file. Basically, don't do that, but if you insist on putting yourself in that position, this might or might not help. Ceki has apparently tested and certifies that prudent mode in logback satisfies his clients requirements for some set of platforms. That would require much more intense testing. We did have a user ([email protected] who filed bug 48704) who was motivated sufficiently to report his issue and explore migrating an application to a different logging framework. It is not far-fetched to believe that he might have been motivated sufficiently to test to see if the patch addressed his issue, improve it and possibly get involved in the community on a regular basis. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
