- From: Peter Karp <pkarp@ai.sri.com>
- Date: Wed, 12 Oct 2005 13:37:39 +0200
- To: Craig Webster <craig@xeriom.net>
- CC: semantic_web@googlegroups.com, Robert Stevens <robert.stevens@manchester.ac.uk>, wangxiao@musc.edu, public-semweb-lifesci@w3.org
Agree furiously... Craig Webster wrote: > Hi, > >>Dear Robert, We certainly do need to enable, but I do see preaching >>as critical as well. I see it happen far too often in our field that >>people jump into large projects and "just do it" without proper >>design and deliberation up front. They end up wasting huge amounts >>of effort, and creating artifacts that waste the time of others >>as well. >> >>Biologists learn the importance of designing their experiments >>carefully; I think it is imperative for the computer scientists >>among us to also teach them the importance of designing their >>software carefully. > > > > While the "just do it" attitude does have problems, there is also a > problem with the "plan it all" attitude. The first gives you a working > but hard to extend artefact. The second give you a beautiful, easy to > extend and maintain artefact which can be hard to work with. > > The model should accurately portray the domain and the best way (in my > opinion) to do this is to walk the middle ground: plan, design, > implement a model which loosely matches the domain and try to use it. After > using teh model and seeing what is lacking you can then extend the model > to match how it will be used. > > Yours, > Craig
Received on Wednesday, 12 October 2005 11:38:05 UTC