Date: Wed, 30 Aug 1995 17:37:43 -0700 From: "Jordan K. Hubbard" <jkh@time.cdrom.com> To: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> Cc: hackers@freebsd.org Subject: Re: /etc/disktab and stuff Message-ID: <7696.809829463@time.cdrom.com> In-Reply-To: Your message of "Wed, 30 Aug 1995 17:04:31 PDT." <199508310004.RAA09965@gndrsh.aac.dev.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> You have now promised a boat load of ``new'' features and plan to implement > them in 2.1 is something that the decision was already made and _AGREED_ > to. > > This is a recuring theme, and yes, I am tearing you a new asshole over > it, no that is not good for the project, so let try to reword this. And I appreciate that. Let me see if I can make my position plainer and see what people on this list think of the whole situation. First, let me be the first to say that this work should have been done MONTHS ago. I know that. For one reason or another, too much stuff happened in my life and the project and I didn't even so much as get the chance to look at it. Such is life, and I wish it were otherwise. Second, let me clarify what the "recurring theme" REALLY is here. The problem is that I hate our install to pieces and consider it as broken as David would, say, a UFS bug that randomly trashed your disk once every other day or so. Would you agree to releasing 2.1 the following day if we discovered such a bug in UFS? Of course you wouldn't - the damage to our reputation would be incalculable and you'd know that as well as everyone else. However, when people can't install FreeBSD or they have a really hard time doing it, it also does damage to our reputation and perhaps you don't appreciate the full magnitude of each shortcoming as much as I do given that such letters are generally sent to _me_. I could forward each and every one to -hackers, and will if people would like me to, but I've just never seen much point in that as I've generally been the only one on the sysinstall hook anyway. If I could put our installation on a timeline, it would look like this: 1.0 - Beyond terrible. There aren't too many words in the english language that go far enough, so let me just summarize the install here as "unacceptably bad by all biological standards" 1.1 - Still terrible. Cosmetic fixes, but terrible, evil, horrible. 1.1.5.1 - See 1.1. 2.0 - First attempt to fix the horror. Addresses many evils, introduces still more. On the whole, a failure. A quick and nasty botch and about as tough and robust as a robin's egg. 2.0.5 - Good intentions, bad, rushed implementation. Rushed because of bad timing and the fact that not enough time was budgeted to deal with slice changes. EVERYTHING in the 2.0.5 install was a trade-off in achieving the absolute minimum in install functionality in the shortest possible time, and the basic approach to installation was subsequently proven to be entirely wrong, like a car with no brakes and the ability to drive on only one road! Would you like to drive such a car? Our users certainly didn't seem to enjoy the experience in the many cases when it crashed them into a wall or didn't allow them to turn just when they really needed to. Again, all code freeze issues aside, 2.1 WAS intended to redress the balance of the many wrongs committed against our user base with 2.0.5 and that was certainly always my understanding during the many conversations I've had over the months with folks like David, Poul and Gary. Be careful what you defend - you just might get it! I'm perfectly willing to just put sysinstall in the can right NOW, as it was in the last snapshot, and that's no threat at all - I really can do that with almost no effort. It should build well enough, and it works well enough for some percentage of users to get from A to Z with it. But only some, and we WILL get a lot of complaints. Trust me, I know what's wrong with that code and you can easily screw yourself very badly with it, assuming that you can even install with it at all. There is a disturbing number of users for whom sysinstall has never worked! I helped them for as long as I could and then regretfully waved goodbye to them as they (with equal regrets) abandoned FreeBSD and went to Linux as the only OS that would install for them. Needless to say, having my name on something that broken causes me almost physical pain, especially when I *know* how to fix a lot of it and know that fixing it will save me and the other project members endless grief in answering the same questions over and over and over and.. Well, you get the picture. The only fly in the ointment is that it's late, and I know that. I am working hard to try to deal with that now, and I DO have a substantial advantage over what I had last time! I have a complete libdisk to work from, I know HOW to work with it now and all the ftp stuff is also ready to grab and fold in. Those two things alone probably accounted for at least 90% of the time invested in the 2.0.5 sysinstall and they, with a few minor tweaks, will work just fine in what I'm working on. I have a proposal. I'll keep working on the system I want to ship, here in my corner with no serious pressure (except for what I apply to myself). If the release date is actually imminent (and it's not at the moment, David's not even done merging bug fixes yet) and I've still not got something better than what we have now, then I'll do what I suggested; I'll just go with sysinstall from 2.1.0-950726-SNAP and be done with it for 2.1. No delays from me, no shifts in paradigm, just the same old thing you all saw from the last SNAP. I'll be sad, but I know that the SNAP worked for as many people (if not more) than 2.0.5, so we'll be no WORSE off at least. As you've resigned from the release team, I don't see the actual methodology we use as affecting your life significantly in any way, just so long as it goes out close to on-time (modulo any other delays beyond my control) and in an installable shape. Heck, since you say you roll your own releases anyway and have for some time, you can even continue using my old sysinstall if you like it better for some reason. I'm certainly not going to personally force you into anything. How's that for a compromise? What do others think? Are perhaps a few weeks in extra hassles worth saving potentially many months of anguised user questions? I think so, but if the rest of you are too eager for 2.1 to even possibly wait that long (and I really am talking worst case, at least from my contribution's side of things) then I'll go with the previous sysinstall. Heck, in some ways it sure would make my life a lot EASIER! I could see it go either way, just so long as I was sure it was what people really wanted. Jordan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?7696.809829463>