The Wood for the Trees blog's Maria Hodges offers her 10 ideas for those researchers looking to build a really "terrible" database: First, make submission of data difficult, because scientists are smart people, and they'll eventually figure out how to do it. And, "who cares if everyone submits to a rival database just because it's easier to use?" Second, have a support service that's only available from Monday to Friday, 9 am to 5 pm GMT, because everyone knows important stuff only happens in Europe during the week. Third, don't let your file formats interconvert, and you should especially ignore the work of the open microscopy environment. Fourth, keep your database independent of others, since you don't necessarily want to be found. Fifth, you should completely trust your automated systems. Sixth, don't provide a unique and permanent identifier. Seven, make sure reviewers can't see the raw data, because "reviewers love to receive emails with thousands of huge images attached." Eight, include an introductory guide of at least 44 pages. Nine, if you include a search option, only make it available in UK English or US English, but certainly not both. And finally, don't bother to develop good visualization tools since scientists love "pages and pages and pages" of data, and making it simple to see connections between datasets would make it just too easy.
Deepak Singh at Business Bytes Genes Molecules agrees with all these suggestions, except number five. Some of the largest data systems in the world are almost completely automated, so it is possible to trust an automated system, he says.