On Fri, 2005-11-18 at 09:45, christian pearce wrote: > As you can see from the response you haven't gotten > (i.e., none) people don't give a lot of thought and consideration to > this.
I wouldn't say people haven't thought about the problem (even more obvious now, than when I started to reply this morning): most of us have had to deal with the situation and the questions that Viraj raised. He seemed to understand the issues on a practical level about as well as anyone could. Sometimes a lack of response just indicates the lack of a real solution. 'Undo' code is by nature one-time, one-off, and becomes meaningless clutter after it's served it's purpose. Cfengine can help us distribute an 'undo', but it isn't what configuration management is typically about, it's just happens to be something you can also do with cfengine. And simple good housekeeping practice suggests deleting it when it's no longer necessary. I'm not sure it should even be considered with other policy or convergence considerations. Undo code can (should) include tests, to exclude the 'new host' or 'already undone host' situation where an undo isn't needed. The question of 'one-time' actions with cfengine has come up before, but not so much in this context of temporary undo's (distinctions being things you would want to do on new builds too, and keep around indefinitely). Mark's suggestion from a year ago for one way to accomplish 'DoOnce' actions could be useful for undo's too: "You can define ifelapsed/expireafter times with a very long time limits: ifelapsed=99999999" ------------ While I'm always interested in hearing about what another tool can do that what I'm using can't do, dissertations purely on what cfengine can't do are mostly academic indulgence. (Don't forget it can't wash the laundry or take the dog for a walk either. Actually, the list is completely open-ended.) Every implementation will teach you something different about the capabilities and limitations of cfengine, but anyone who works with cfengine every day has a very practical understanding of what it's good for, and where it's not so good. For all the talk about "solutions" in the clarification of the pdf linked to earlier this morning, I didn't see any at all, in either the pdf or the 'clarification'. -Ed "If you meet the "gold server" on the road, kill it!" _______________________________________________ Help-cfengine mailing list Help-cfengine@gnu.org http://lists.gnu.org/mailman/listinfo/help-cfengine
