> My hunch is that the memcache devs have probably > considered this idea but never went ahead with > it..... I would be interested to know why?
I'm not one of the devs, but I can guess why they wouldn't want to do this. It's pretty simple: it's not just expiration and expulsion that can cause a memcache node to lose keys. If you need to reboot a server, the data will be lost. If you have a hard disk crash and the server dies, you lose your data. Of course, you can reboot database servers and so on as well, but they're usually replicated so that there's no interruption of service and no data lost. Besides, this feature could affect performance, and memcache is ALL about performance. For us, our session data is usually mostly static. When you log on, we create the session data and it doesn't change much until you log off. So we actually store it in the database AND in memcache. The database then is mostly write-only and only needed if a memcache node goes down and the key is lost. Dean.
