From owner-freebsd-current Fri Dec 5 12:21:20 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id MAA03780 for current-outgoing; Fri, 5 Dec 1997 12:21:20 -0800 (PST) (envelope-from owner-freebsd-current) Received: from tera.com (tera.tera.com [207.108.223.21]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id MAA03751; Fri, 5 Dec 1997 12:21:04 -0800 (PST) (envelope-from kline@tao.thought.org) Received: (from uucp@localhost) by tera.com (8.8.7/8.8.7) with UUCP id MAA03154; Fri, 5 Dec 1997 12:19:41 -0800 (PST) Received: (from kline@localhost) by tao.thought.org (8.8.5/8.7.3) id MAA11678; Fri, 5 Dec 1997 12:06:00 -0800 (PST) From: Gary Kline Message-Id: <199712052006.MAA11678@tao.thought.org> Subject: Re: 3.0 -release ? In-Reply-To: <28074.881287489@time.cdrom.com> from "Jordan K. Hubbard" at "Dec 4, 97 06:04:49 pm" To: ports@freebsd.org Date: Fri, 5 Dec 1997 12:06:00 -0800 (PST) Cc: asami@cs.berkeley.edu, toasty@home.dragondata.com, nate@mt.sri.com, jak@cetlink.net, current@freebsd.org, ports@freebsd.org Organization: <> thought.org: public access uNix in service... <> X-Mailer: ELM [version 2.4ME+ PL32 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk According to Jordan K. Hubbard: > > > Regarding ports and packages in general, I think that we also need to > start looking *really seriously* at the idea of a "multi-volume" > ports/packages collection, since a need to split them across at least > 2 locations has already been the case since 2.2.5. Even if you prune > and chop and tweak like mad for 3.0, I can guarantee you that the > packages won't all fit in 650MB [see note]. :-) Let me run some thoughts and questions up the flagpole, Jordan. We could try these for a year or so since there really isn't any easy way of market-researching this. Two types of FreeBSD RELEASE: one would be a 4- to N-CD release at cost+margin. $50 or 50+. This no more than twice a year; these releases would be as before with everything available. The second type of RELEASE would probably fit onto one CD, the kernel, compilers, source, X11R6 bins. The ports would be in a separate CD package, but source-only. What's the reason for having binary distfiles when--given the proper *.mk files-- everything can be built from the source? > > This means to me that the INDEX file needs to grow at least one more > field for the volume name (or maybe we can just tuck it into the > keywords field, does anyone even use that? :-) and somebody needs to > modify portlint so that a port's volume is checked against its > dependencies, it being an error to put your package outside the same > volume as all the things it depends on. This would allow me to modify > sysinstall to request in turn that the appropriate media be mounted > when asking for a specific package that isn't on the current media. Yes! This is why I pull things off the net _only_. > > To think about: Should we also implement a volume-to-media mapping > file which allows the installation to build a menu of valid choices > based on the media types the user has available and the "map" of where > the packages are available, be that anything from "CD:WC/3" to > "ftp://ftp.jp.freebsd.org/pub/FreeBSD/packages/japan-latest"? > Thinking ahead to where we've got potentially 4000-5000 packages, with > an average size of 500K for packages the whole collection is going to > be easily 2GB or more. I could see where it might become distributed > more geographically, or from a more limited number of servers than > supply the distribution bits, and we're going to need to account > for that somehow. It's time to consider this point in some depth. It is fantastic to see us growing internationally. As time passes, most of the programs will probably originate in the non-English, non-Latinate corners of the globe and have to be ported from, say, Chinese to English. The non-English sites would carry packages in their language. With links to other geographic sites. By the time we are this diverse, hopefully the DVD drives will solve the space|packing problem so that a RELEASE will again fit onto 1 or 2 platters. > > Doing this would also allow Walnut Creek CDROM to see just how much > revenue the ports collection alone generated, and as a new product I'm > pretty confident that I could negotiate in advance that a slice of the > pie from every ports collection CD sale go into a special "support the > ports collection project fund" (over and above money already given to > FreeBSD.org) from which we could buy Satoshi and crew a package > building machine from hell, among other periodic goodies. :-) [and > I've been thinking of some nice 10 drive CCD array configurations > which would really do the job nicely when mated with a dual-processor > PII-300. ;)] What would it take to get a ports platform up. With a T1 line? Could we do build+test on that box as well as our own? I've run into unfathomable built problems with several ports retrieved from freefall/ports and have given up? Having one unified site to work with would give us the opportunity to do a port right. Obviate lots of head-banging and digging into build problems. These last two paragraphs are off-topic. Feedback on the above, people?? gary > > What do folks think of all this? > > Jordan > > -- Gary D. Kline kline@tao.thought.org Public service uNix