If they were active they'd be rediscovered again as you noted, but there's lots of useful information that you'd lose if you did that, so it's definitely better to have some automated way of at least cutting down the list of machines to the ones that are really likely to be out of use. This has come up before and it's something that isn't unfortunately able to be directly dealt with in GSS as it is; there isn't a timestamp field in the database, and it's not possible to add one easily (because of the edition of Sybase we were using, database changes tended to have to be done with major releases).
I've been thinking about what I could possibly do as a script which could take a guess at the date and delete the clients. There are basically four useful sources of information a script could use:
- the date on the file containing the machine default configuration data for each client, which is updated during a task with a "Configuration Refresh" step
- the date on the file containing the last retrieved WMI Inventory data for each client, which is updated during tasks with an "Inventory Refresh" step.
- the dates in the Task Log stored in the database, which can be correlated with the clients that successfully ran the task (this is a bit more complex to do).
- scraping the information about client messages stored in the Configuration Server's ngserver.log file (this is a bit more complex again).
Depending on how the console is used, it may be that even doing only the first two should give a reasonable first approximation of which clients are suitable candidates for removal (you can get a sense of that by having a look at the "DB\DATA" subdirectory inside the console installation, where the configuration and inventory files for each client live, and sorting the folder by date).
I've got a lot of other work on at the moment, but if you create an enhancement request at the
script issue tracker, then if I can come up with a script which can use the above information to do a date-filtered bulk delete then that's the best way to go about it (and anyone else interested in this can star the issue too).
By the way, one small bit of trivia; the actual first version of the Ghost console was something we started working on almost immediately after Symantec acquired Ghost from Binary Research, and its initial release (the real 1.0 of that code) actually was in Ghost Enterprise 6.0, released mid-1999-ish. It didn't really get much attention back then, and it didn't really gain enough critical feature mass for a while, but the core of it has been around a long time.