Date: Thu, 5 Feb 1998 07:15:10 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: jkh@time.cdrom.com (Jordan K. Hubbard) Cc: julian@whistle.com, hasty@rah.star-gate.com, mckusick@McKusick.COM, hackers@FreeBSD.ORG Subject: Re: kirk's soft-update integration.. Message-ID: <199802050715.AAA09623@usr08.primenet.com> In-Reply-To: <240.886636132@gringo.cdrom.com> from "Jordan K. Hubbard" at Feb 4, 98 03:48:52 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> Ummm. I still completely fail to see why OpenBSD was able to > integrate all the hooks AND make the two "encumbered" (sorry > to use that word ;) files available on their web site without > any such hoop-jumping. Because OpenBSD has not deviated so significantly from BSD4.4-Lite2 as FreeBSD has? There are interactions with the vnode changes in FreeBSD, the cache unification (Kirk's original code keeps two lists of vnodes), and the VM system. > I really do also get the feeling that the intervention of whistle in > this matter has only vastly overcomplicated the situation for the > average user. As amancio says, why can't we just ftp the files from > someplace? IMO, not speaking for Whistle here, you can. > I've asked Kirk about the method that OpenBSD used and he > didn't seem to have any objections to what they're doing, so why will > FreeBSD's support require a signature in blood before we can do the > same thing? It won't, IMO, again not Whistle's. I think the question being asked by Julian is about integration of the hooks, despite the fact that commercial use of what they would hook requires obtaining a license from Kirk, and non-commercial use requires source distribution, ala Sean's GPL reference. I believe it's on the order of the GPL'ed math emulator's hooks. Don't forget, that non-licensed-in-this-way patches and an improved "updated" (renamed "syncer") that effectively does write-gathering (ala SunOS's big performance improvement for their NFS servers) are packaged with the stub version of the update code (which *is* usable commercially. Personally, I think it's a win, even if I might have chosen to eat the recalculation of Hamilton cycles (as a trade for allowing dependencies to cross FS stacking boundries) and a more general soloution to the problem. Such a soloution might not have suited Whistle, actually, since it would have a slightly lower performance (3-5%, probably) and a significantly higher mount and crash recovery overhead (you don't mount or crash-recover that frequently). I've seen it "alpha" with Kirk's live dependency code, and I have to say that it's damned impressive to see something like this work in FreeBSD (I saw it work in Windows95, without clustering, about 2 years ago when my team did it at Artisoft; it was less impressive that running it in FreeBSD ;-)). Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199802050715.AAA09623>