| |||||||||||||
OpenMRS might help more than the developing worldOpenMRS might help more than the developing worldPosted Jun 21, 2007 16:34 UTC (Thu) by shane (subscriber, #3335)In reply to: OpenMRS might help more than the developing world by dion Parent article: Google Summer of Code Series, OpenMRS using the typically catastrophic waterfall model of software development Somewhat off-topic, but the waterfall model gets a lot of bad press. However, in a properly managed software engineering environment, the only downside is cost. The aerospace industry uses the waterfall model exclusively, because it is the only way to get software that is proven to work to the level of safety and reliability required. I'm not saying the Danish journal system is not a huge scandal, but don't blame a particular software development methodology.:)
(Log in to post comments)
OpenMRS might help more than the developing world Posted Jun 21, 2007 18:08 UTC (Thu) by dion (subscriber, #2764) [Link] Well, the waterfall model assumes that you know everything about what you are doing before you start doing it.
So it might work well for very strictly defined problems, like "fly space shuttle", if you do it right.
In the hands of dimwitted, blanco-check writing bureaucrats it's an absolute unmitigated disaster.
Any problem that leads to discoveries during implementation (read: almost all of them) you need something less rigid and less risky.
We as a country has a long history of huge software projects that turn out to be DOA, I remember one system (AMANDA) that took years and millions to produce that ended up being scrapped because it depended on OS/2 and one specific BIOS version, long after both had been EOL'ed.
I think that it would be a great damage control strategy for the government to say that the largest payment that can be made on any governmental software project is 200k DKK (or somesuch) and that the software must be usable before any payment can be made.
A hard cap on single payments will force projects to be cut into smaller pieces and any disaster will be limited to each small sprint rather than be allowed to fester and drag the entire GNP down with it.
That way we could avoid these insane budget overruns and huge software disasters.
It doesn't make any sense to build software like you build bridges, a half finished piece of software can be plenty useful, if not for the end users then for the designers and administrators.
If I was in power I'd also mandate that only my own svn.gov.dk could be used for source control and that all software developed must be publicly accessible and under a Free license.
I'm pretty sure that mandating Agile and OSS would bring down the costs and the risks in these huge administrative software projects.
It's a shame that our politicians are universally IT-retarded so they fall for every trick in the contractors want to pull.
waterfall vs. carousel Posted Jun 22, 2007 4:03 UTC (Fri) by xoddam (subscriber, #2322) [Link] > A hard cap on single payments will force projects to be cut into> smaller pieces and any disaster will be limited to each small > sprint rather than be allowed to fester and drag the entire > GNP down with it.
So what you're saying is "release early, release often", yes?
OpenMRS might help more than the developing world Posted Jun 22, 2007 10:46 UTC (Fri) by nix (subscriber, #2304) [Link] Are you sure you don't write for the Private Eye? The writing style (and litany of entirely justified complaints) seems familiar somehow ;}
OpenMRS might help more than the developing world Posted Jun 22, 2007 12:28 UTC (Fri) by dion (subscriber, #2764) [Link] Am I supposed to be flattered or offended by that?
flattered or offended Posted Jun 22, 2007 15:03 UTC (Fri) by copsewood (subscriber, #199) [Link] Put it this way, if something I had written had been compared favourably to that long running, satirical, frequently sued, perceptive and witty UK rag I would have felt highly flattered, but not everyone would.
flattered or offended Posted Jun 23, 2007 21:17 UTC (Sat) by nix (subscriber, #2304) [Link] It was meant to be flattering. :)
Shuttle iterative development Posted Jun 22, 2007 17:51 UTC (Fri) by man_ls (subscriber, #15091) [Link] So it might work well for very strictly defined problems, like "fly space shuttle", if you do it right.As you can see from the program managers, it doesn't work [pdf]. An excerpt: Shuttle program schedules required that the software be certified and ready to support the first orbital flight. However, the detailed definition of all requirements could not be completed in time to support a proven software design, implementation, and verification development cycle (Figure 2) due to the ongoing vehicle engineering analysis work. [...] To satisfy these conflicting demands and still deliver a fully verified, error-free software system consistent with Shuttle flight schedules, a development strategy was evolved that preserved the effectiveness of the proven development cycle and satisfied the customer requirements.
OpenMRS might help more than the developing world Posted Jun 22, 2007 17:43 UTC (Fri) by man_ls (subscriber, #15091) [Link] The aerospace industry uses the waterfall model exclusively, because it is the only way to get software that is proven to work to the level of safety and reliability required.Where do you get that idea? The Space Shuttle primary flight software was developed using an iterative model: they put out releases every few months, not unlike the current kernel development process. It's even mentioned as an example on the wikipedia article. The pure waterfall is definitely not the only way to get reliable software, in fact it is often the worst. Why not blame the methodology if it is not right for the job?
OpenMRS might help more than the developing world Posted Jun 25, 2007 9:35 UTC (Mon) by shane (subscriber, #3335) [Link] I got it from a talk at the 23rd Chaos Communication Congress last December, Software Reliability in Aerospace, given by a software engineer who works in aerospace.The aerospace industry uses the waterfall model exclusively, because it is the only way to get software that is proven to work to the level of safety and reliability required.Where do you get that idea? If you think about it, the waterfall model makes a great deal of sense for aerospace projects. In these, the requirements are usually understood in much more detail before coding begins than for a lot of software projects. Also, while getting the requirements wrong is costly in any project, when you are working with very expensive hardware it is disastrous. Finally, the benefit of having a partially-working system (for example just "core features") is much less or perhaps even nonexistant for aerospace.
DO-178B and waterfalls Posted Jun 25, 2007 21:19 UTC (Mon) by man_ls (subscriber, #15091) [Link] Fascinating stuff, thanks!
|
|||||||||||||