Position Paper by Betty Salzberg, Northeastern University,


Academic Researchers need more information from industry to do research which has impact.

Most of the people doing research in universities in the area of database systems would like to publish ideas which could be used in new products, to make old products more efficient; or to suggest new features which could be added to products. The time-line for these innovations could be longer than that in industry, but they should be "realistic" or "practical." However, most academics do not know enough about how commercial products work to do this.

A good example of a lack of relevant information is in Jeff Vitter's 1985 TODS article on WORM optical disks, which could never be used on optical disks, because WORM optical disks have a smallest write unit of 1000 bytes with a 300 byte checksum burned in the disk to make up for the high error rate. Vitter's article assumed a few bits could be changed in a write operation so that empty space at the end of a block could be used for updates. Vitter did not know enough about optical disks when he wrote that article. (In addition, since the article is in the top journal, this lack of knowledge is passed on to people who assume the top journals and conferences have reliable information.)

One way to propagate information about how products really work would be for academics to visit industry or collaborate with industry people to analyze ideas and describe algorithms already being used in commercial products. This would be really useful for the research community and might enable academics to get the foundation they need to do meaningful research. This would require some cooperation from industry. Examples of descriptions of algorithms in use appear in the ARIES papers (in addition to the original ideas, system R and DB2 mechanisms were described in detail) and in the Gray/Reuter textbook. For academics to write this kind of material, they need to have access to information of this sort and they do not have this.

When we invited people from industry to talk in a recent NSF workshop (slides from the talks are at www.ccs.neu.edu/groups/IEEE/ind-acad/), we got a lot of ideas of topics we academics could work on. Most of these had to do with managing or modelling complex systems. To do this, we would have to know about how these systems work. I don't know how to start to find this out. We can't do research you can use and we can't teach students what they need to know to do good work unless we can get more good information. Most academics would prefer to write papers which are useful and interesting to people in industry. But we need help.