Date: Fri, 06 Jun 2003 01:16:45 -0700 From: Terry Lambert <tlambert2@mindspring.com> To: "Matthew D. Fuller" <fullermd@over-yonder.net> Cc: arch@freebsd.org Subject: Re: Making a dynamically-linked root Message-ID: <3EE04DED.DCC1E637@mindspring.com> References: <20030603113927.I71313@cvs.imp.ch> <16092.35144.948752.554975@grasshopper.cs.duke.edu> <20030603115432.EGLB13328.out002.verizon.net@kokeb.ambesa.net> <20030603122226.BGPM11703.pop018.verizon.net@kokeb.ambesa.net> <3EDD81A4.B6F83135@mindspring.com> <3EDDF732.1060606@tcoip.com.br> <3EDF2B1C.6E9C892E@mindspring.com> <20030605221114.GB51432@over-yonder.net> <3EE04642.602DA5EF@mindspring.com> <20030606080232.GE51432@over-yonder.net>
next in thread | previous in thread | raw e-mail | index | archive | help
"Matthew D. Fuller" wrote: > On Fri, Jun 06, 2003 at 12:44:02AM -0700 I heard the voice of > Terry Lambert, and lo! it spake thus: > > > > I'd argue that it's lost in the noise. > > I'd agree. > > I think what you're meaning (though it's not quite coming across) is that > for most people and most situations, the downside is either nonexistent, > or so small as to be so for all practical purposes. I'm more or less trying to play "devil's advocate" on this. There are so many commercial systems that dynamically link everything, that I have to believe that their highly paid engineering resources thought it through before flipping the switch. > On the other hand, there are people and situations who believe that the > [potential] downsides are much larger. We differ in that I don't believe > that all such claims are bogus. Sure, many of the examples put forth are > rather arbitrary and contrived, but I've found that I literally can't > come up with a scenario so arbitrary and contrived that SOMEBODY isn't > dealing with pow(it,N) every day. And I'd agree with this. I'm sure Peter Wemm, who posted earlier, is dealing with a large number of serial console based systems, and occasionally has a system that doesn't have a hot spare, or has data on it that the normal approach of reinstalling on new hardware and replacing it won't work for him. With the number of machines he has, it's potentially a big deal. I also know that a former employer didn't have any local boot media available on their devices, and only a serial console; you either were very careful, or you took out a zillion screws to get at the only media it had and/or the floppy radial connector. I would class these systems as "fragile", but I have no doubt other people have followed the same model. At a former former employer (Whistle), the fragility problem was resolved using two boot partitions and "bootnext"; this same thing is available to those people with fragile systems, though it's not completely convenient, mostly because FreeBSD / and /usr partitions aren't easily rearrangeable to be mounted read-only. I think after you last posting, though, that at least you and I are at a consensus: set it up to work both ways, for now, and flip a big red switch on the default in the 5.3/5.4/6.0 time frame, with plenty of advance warning. If people want it the other way, then they can flip the switch the other way, and do a "make world" and burn their own ISOs (or hack the install and build code to be able to install either way). Frankly, I think that no one is going to miss the static linkages, when the switch is finally thrown. -- Terry
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3EE04DED.DCC1E637>