Hi! > > > If we go the "STD = STR plus image-to-disk" route, > > > ... > > > > Well, STD != STR plus image-to-disk. > > > > For std you need to free 50% memory, plus you need freezing phase, > > plus atomic snapshot. That's pretty different code. > > The "free 50%" is an artifact of how snapshotting is being > done just now, right? Anything that's not modified during > the snapshot (like the kernel text segment) doesn't need > to be copied. I suspect someone wanting to be clever could > map most memory as read-only and kick in some copy-on-write > logic to reduce that requirement by quite a lot ... :) > > Anyway, that detail aside: image-to-disk is exactly > creating that snapshot (50% or otherwise) and doing that > atomic thing ... and STR needs freezing too. So you've > not convinced me that, within the limits of a soundbite, > it's misleading to say that "STD == STR + image_to_disk". STD is quite complex process. Yes, you can say STD == STR + image_to_disk, but STR == "inform drivers what is going on" in this case... Informing drivers is the only thing STR and STD have in common (stopping processes can be viewed as part of telling drivers, and BenH avoids doing stop processes on ppc). Pavel -- People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
