Date: Tue, 1 Apr 1997 15:27:42 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: jkh@time.cdrom.com (Jordan K. Hubbard) Cc: proff@suburbia.net, terry@lambert.org, joe@pavilion.net, gbeach@cybernet.com, hackers@freebsd.org Subject: Re: Internal clock Message-ID: <199704012227.PAA12232@phaeton.artisoft.com> In-Reply-To: <27313.859933305@time.cdrom.com> from "Jordan K. Hubbard" at Apr 1, 97 02:21:45 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > The sad reality is if these things are not incorporated in -current > > then they fall by as the original authors move onto other projects. > > The problem is that we need more committer/reviewers who are capable > of bringing stuff in at this scale without also breaking everything. > That's not as easy as it sounds, and even some of our best current > people still break the tree now. :-) > > Breaking the tree is also a problem because of the amount of tech > support it generates. It might not bother the end-user to have to > stay away from -current for a few days, but as one of the developers > who gets subjected to an endless stream (many messages a day!) of mail > saying "current is broken!" "did you know current is broken?" "hey, I > just thought I'd let you know that current is broken again. Get your > act togther, guys!" every time current is broken, I mind it very very > much. Break the tree for 3 days and you've just guaranteed about 200 > messages in my inbox (sent to me *directly*, not even through the > mailing lists - those account for another 300 or so). > > If people have begun to wonder at our conservatism, that's one very > big reason for it. Do we need to go back to the discussion how a primitive transaction model can make this work and make all your detractors have nothing to detract about? Even at 10% overhead (to be overzealous on the down side) is worth it if you can safely accommodate 11.11% or more additional committers! STATUS: TREE IS BUILDABLE, SELF CONSISTANT BEGIN COMMIT PROCESS CHECK OUT TREE APPLY CHANGES TO CHECKED OUT TREE <---------. BUILD TREE | TREE SUCCESSFULLY BUILT? ------------ no ---' CHECK IN MODIFIED FILES END COMMIT PROCESS STATUS: TREE IS BUILDABLE, SELF CONSISTANT Are you worried about the CVSSup'able tree being buildable at all times? Or CTM delta's? Don't operate against the active tree, operate against a copy, and only make the copy when there isn't a commit process in progress (Implement type non-specific locking). Are you worried about the space for another copy of the tree? Implement mulitple reader/writer locking so you can use the same tree. Are you worried about two committers going at it at the same time? Implement single writer locking against the tree to keep them from stepping on each other's toes, and which the copy/sup participates in. Are you worried about all of these issues simultaneously? Are you worried about allowing concurrent readers? Implement multiple reader/single writer locking. Regards, 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?199704012227.PAA12232>