Rebuild-in-place

Post #32,757 by kmself 3/19/02 3:02:30 PM Reply	Rebuild-in-place There's a distinction between a full tear-down, and a comprehensive, incremental, rebuild. This is Spolsky's distinction, as I understand it. The parts of the structure you aren't rebuilding provide context for those you are. The "rebuild from scratch" scenario loses this structure. You see a similar move in architecture -- I recall several buildings at UC Davis which were largely rebuilt from the inside out. Expensive, but classic facades were retained as an overall win. Frankly, there are cases I'd go one way or the other. I do think that the "learn from your mistakes" message is a valuable one, and is the core of his argument. `-- Karsten M. Self [link\|mailto:kmself@ix.netcom.com\|kmself@ix.netcom.com] [link\|http://kmself.ix.netcom.com/\|[link\|http://kmself.ix.netcom.com/\|http://kmself.ix.netcom.com/]] What part of "gestalt" don't you understand?`
Post #32,766 by ben_tilly 3/19/02 4:00:18 PM Reply	And that is where my disagreement with him is Frankly, there are cases I'd go one way or the other. Ditto. And I have rules of thumb for when I would do so. And why. Moreover I can point to case examples and research for some of those rules. But he doesn't acknowledge that there are any justifiable rewrites. Ever. No matter what. What should I expect? He is an old Microsoftie. Things must always work best the one way that he saw them work well. Feh, Ben "... I couldn't see how anyone could be educated by this self-propagating system in which people pass exams, teach others to pass exams, but nobody knows anything." --Richard Feynman
Post #32,779 by wharris2 3/19/02 5:35:21 PM Reply	I still think he's being inconsistent Okay, so he carefully changed the code so that everything still worked from build to built, from checkin to checkin. He's kept the functional requirements the same, but he's still done a lot more than basic refactoring, having rewritten a lot of it from scratch. Where each demon is slain, more hate is raised, yet hate unchecked also multiplies. - L. E. Modesitt
Post #32,807 by JayMehaffey 3/19/02 10:03:57 PM Reply	The difference is The difference is that he's still stuck with all the basic architectural flaws of the orginal design. Jay
Post #32,829 by kmself 3/20/02 4:48:52 AM Reply	Nail I think that's really it. If you're working with fundamentally sound architecture (software or structural steel, I don't care), then you can recover from a shitty or time-worn implementation. If the architecture itself is shot, then going from scratch is pretty much your only choice. There are classic old buildings (again, good fundamental architecture) I've seen rebuilt. More impressively was a company that operated near where I used to go to school. Tracor Aviation builds jets. The jets they build have numbers like 737, 777, and even (for small values -- the short stubby birds), a 747. What they do is rebuild a Boeing in place, from the ground up. The airframe is totally replaced, but it's the same bird. ...the point being that that's cost-effective. Which is what this all really comes down to. `-- Karsten M. Self [link\|mailto:kmself@ix.netcom.com\|kmself@ix.netcom.com] [link\|http://kmself.ix.netcom.com/\|[link\|http://kmself.ix.netcom.com/\|http://kmself.ix.netcom.com/]] What part of "gestalt" don't you understand?`
Post #32,832 by ben_tilly 3/20/02 7:23:16 AM Reply	I think he is consistent The act of writing software from scratch is a very different act from transforming existing code. Different steps, a different feel to the work, and different results. He thinks it is always a better way to work. I think it is good to know about but only sometimes better. Cheers, Ben "... I couldn't see how anyone could be educated by this self-propagating system in which people pass exams, teach others to pass exams, but nobody knows anything." --Richard Feynman

Welcome to IWETHEY!