3 ms·
He beat me to it as I'm mentally working on a similar essay. Backward compatibility and shipping pressure I already covered a lot in my posts on Schneier's blog
by nickpsecurity 11y ago
He beat me to it as I'm mentally working on a similar essay. Backward compatibility and shipping pressure I already covered a lot in my posts on Schneier's blog elaborating on this. See Steve Lipner's Ethics of Perfection essay for a great take on "ship first, fix later" mentality. He had previously done a high-assurance, secure VMM. So, he had been on both sides.
https://web.archive.org/web/20150319122817/http://blogs.microsoft.com/cybertrust/2007/08/23/the-ethics-of-perfection/ https://web.archive.org/web/20150319122817/http://blogs.micr...
On backward compatibility, you need to explore lock-in and network effects. These are the strongest drivers of the revenues of the biggest tech firms. Once you get the market with shipping, people will start building on top of and around your solution. They get stuck with it after they do that enough to make it hard to move. Familiarity with language or platform matters here, too. The economics become more monopolistic where you determine just enough additions to keep them from moving.
I agree with a commenter there that it needs a OpenVMS tie-in: a great example of Right Thing vs Worse is Better that won in market. While their management was good. ;) It had better security architecture, individual servers went years without reboot, mainframe-like features (eg batch & transactions), cross-language development of apps, clustering in 1980's, more English-like command language, management tech, something like email... the whole kitchen sink all integrated & consistent. Reason was it was a company of engineers making what they themselves would like to use then selling it to others to sustain it. Also mandated quality where they'd develop for a week, run tests over weekend, fix problems for a week, and repeat. That's why sysadmins forgot how to reboot them sometimes. ;)
https://en.wikipedia.org/wiki/OpenVMS https://en.wikipedia.org/wiki/OpenVMS
Here's a few others that fall under Cathedral and Right Thing model that got great results with vastly fewer people than Worse is Better and/or were successful in the market. Burroughs and System/38 still exist as Unisys MCP and IBM i respectively. Lilith/Oberon tradition of safe, easy-to-analyze, and still fast lives on in Go language designed to recreate it. There's nothing like Genera anymore but Franz Allegro CL still has a consistent, do-about-anything experience. QNX deserves mention since it's a Cathedral counter to UNIX where they implemented POSIX OS with real-time predictability, fault-isolation via microkernel, self-healing capabilities, and still very fast. Still sold commercially and was how Blackberry Playbook smashed iPad in comparisons I saw. They once put a whole desktop (w/ GUI & browser) on a floppy with it. Throw in BeOS demo showing what its great concurrency architecture could do for desktops. Remember this was mid-1990's, mentally compare to your Win95 (or Linux lol) experience, and let your jaw drop. Mac OS X, due to Nextstep, could probably be called a Cathedral or Right Thing that made it in market, too.
http://www.smecc.org/The%20Architecture%20%20of%20the%20Burroughs%20B-5000.htm http://www.smecc.org/The%20Architecture%20%20of%20the%20Burr...
https://homes.cs.washington.edu/~levy/capabook/Chapter8.pdf https://homes.cs.washington.edu/~levy/capabook/Chapter8.pdf
https://en.wikipedia.org/wiki/Lilith_%28computer%29 https://en.wikipedia.org/wiki/Lilith_%28computer%29
http://www.symbolics-dks.com/Genera-why-1.htm http://www.symbolics-dks.com/Genera-why-1.htm
http://www.qnx.com/products/neutrino-rtos/neutrino-rtos.html#technology http://www.qnx.com/products/neutrino-rtos/neutrino-rtos.html...
https://youtu.be/cjriSNgFHsM?t=16m5s https://youtu.be/cjriSNgFHsM?t=16m5s
So, more food for thought. The thing the long-term winners had in common is that (a) they grabbed a market, (b) they held it long enough for legacy code/user-base to build, (c) incrementally added what people wanted, and (d) stick around due to legacy effect from there. Seems to be the only proven model. It can be The Right Thing or Worse is Better so long as it has those components. So, we Right Thing lovers can continue to trying to make the world look more Right. :)