Sign in or Join FriendFeed
FriendFeed is the easiest way to share online. Learn more »
Deepak Singh
When “good enough” just doesn’t cut it -
Is this a symptom of writing software for publication and then moving on? - Michael Barton
It's a symptom surely of what the measured endpoint is - getting the data out for the paper - not producing something that has real utility. It's the good enough _for what_ bit that is the problem here surely? - Cameron Neylon
That's fair - constant suprise to me that I seem to know more about software development best practice than the academic researchers I talk to. I blame Greg Wilson of course...need to get Software Carpentry or similar course made compulsory for all science undergraduates :-) - Cameron Neylon
Or write it into the grant conditions. That spells out it out pretty clearly... - Cameron Neylon
Neil, I agree with you. I think that's going to change, as more and more software developers enter the life sciences, folks who care about maintenance, quality, etc. But the PIs are still a problem. Of course, this is not just academic research though. I've seen it in companies and perhaps that's the difference between between someone who stays middle of the road and someone (someone could be an entity) who excels - Deepak Singh
couple of off topic things: I think your RSS is not working, or you might have changed it. It doesn't show up on my GReader. Also your Fork Me link to GitHub is not pointing to your account. - pn
Paulo, the feed seems to be OK at this end. and yep, do need to fix that Fork Me link. Thanks - Deepak Singh
Survival of the fittest will show how good things are. In the (free) open source world quality/time_to_invest will show, and for commercial world the quality/price will do the same. - joergkurtwegner
Joerg, I think that's beginning to happen, especially with open source alternatives pushing purchasing behavior. Plus expectations have changed. No one is going to use an internal search engine with a several millisecond response time, when you are used to Google - Deepak Singh
Perhaps I'm the pessimist, but if all scientific software were merely 'good enough' I'd be in heaven. Good enough would at least imply that it compiles/runs/etc. - Paul J. Davis
I think "good enough" in software is favored when an individual needs to get something done and faces limitations in terms of time or financial resources in accomplishing the task. Within those constraints, "good enough" is the best way of making progress rather than waiting 'til someone writes the best possible code. It shouldn't remain that way, but if its cutting edge research, a clear market demand may not have been established as an incentive for some one to create a particular piece of software. - Jill O'Neill
Jill, I've seen enough evidence where that's not the case. PI's tend to lose interest when they have papers published, or if a grad student or postdoc leaves. In the case of commercial entities, it's a cultural thing. Constraints can lead to phenomenal code. - Deepak Singh
Isn't it similar to the evolutionary selection, with academics having a set of "pressures" different from those needed to develop #1-type software? Once the paper gets published, there is no pressure for researcher to improve the code, and things remain "good enough". While in a commercial setting there is always strong pressure from the side of the customer/competition, which drives the development further. I.e. to solve the problem one needs to bring some kind of pressure element to the academic setting. - Yaroslav Nikolaev