Date: Sun, 25 Jun 1995 22:45:01 -0600 From: Warner Losh <imp@village.org> To: "Jordan K. Hubbard" <jkh@FreeBSD.ORG> Cc: hackers@FreeBSD.ORG Subject: Re: Announcing 2.0.5-950622-SNAP Message-ID: <199506260445.WAA04962@rover.village.org> In-Reply-To: Your message of Mon, 26 Jun 1995 01:06:42 BST
next in thread | raw e-mail | index | archive | help
: If you can think of a way of making dual-boot work in all possible : scenarios, then I'm certainly not adverse.. Hmmm, more think real hard :-) I will have to do that. A question I'm not qualified to answer: Which would be harder: Hacking the boot code to do the right thing, or hacking the file system code to allow mounting of subtrees on mount points. Something like you get with nfs: mount fred:/usr/baz /usr If you could do that, then you could specify which part you want to boot. Or you could mount what would be the /usr partition on /usr-2.0.5R and have a synlink from /usr. Likewise with /etc, /var, etc. You'd likely have to make the kernel be smart enough to suffle the symlinks, or you'd have to do it by hand in /etc/rc. There is a drawback with things like /usr/local, but maybe that could happen. Alternatively, you could hack all the shells to read the default path from a known location, and setup things this way, but that doesn't solve the config file problem (which, if they are all in /var, wouldn't be a problem since you'd have different /var's). An idea would be sym links from something like /etc/fstab to /etc/fstab-2.0.5R. I'm not happy with this idea, however, since it seems to involve a lot of kruft. Windows NT does a variation on this, and VMS does a similar thing with logical names. Root would remain the same, and you'd boot kernel.2.0.5R or kernel.SNAP-950622 or something like that. Oh, if I went the massively symlink'd route, it would be ONLY for people that wanted to have n copies of the OS on the machine. If you wanted only one, then you'd not have the performance hit of the symlinks. The two disk case is easy, so I won't talk about it much (I have it mostly setup now, but not automatically). Anyway, are enough hackers interested in this discussion, or is it time to create a list? I would have thought some of my comments would have had more discussion than I've seen in my personal mailbox and in -hackers so far. I'd also like to talk about automatic tools for the generation of the deltas and things like that which would be needed for an upgrade framework to be successful. Comments? Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199506260445.WAA04962>