Monday, July 24, 2006

Brats, bibliographies, peer review

Two things that caught my eye this week:

Lamentations on user interfaces. In the July+August issue of interactions the editor has a rant about "woe-besieged" users. He says

Office productivity tools make you do somersaults to undo their automatic formatting. .. When you try to do something that a product's designers didn't anticipate, some products exact their revenge. They do the digital equivalent of throwing everything into a heap, leaving it up to you to clean up the mess all by yourself. This is marginally acceptable behavior from a two-year-old child...

Also this week I read an insightful rant by a writer for MIT Technology Review talking about Web 2.0. The part that I thought was interesting was when he says that most users need very little of the functionality provided by software manufacturers. This might remind you of the 90-10 (or 80-20, or ...) heuristic.

Here's the relevant quote:

For years, software makers, notably Microsoft, have struggled with the bloatware dilemma. A small fraction of their users want specialized, elaborate new functions; moreover, the software makers themselves need to keep adding features to justify upgrades. But the more niche features they add, the more complex, buggy, and expensive their programs become, and the more off-putting they can seem to most users. The likes of Voo2do, iOutliner, Google Calendar, and the new Google Spreadsheets have solved this problem by ignoring it. They do most things that most users of their desktop counterparts want -- but almost nothing that the specialized user might. Writely lets me make bullet-point lists and choose from several fonts -- but I can't add footnotes or easily change the column layout. Google Spreadsheets lets me enter formulas and values as easily as Excel does, but it cannot produce graphs or charts. And the online to-do list systems lack some of the more sophisticated features I like in BrainStorm and Zoot. The result of this short-tailism might be a curious new "long-tail" division between online and desktop applications: the free online apps will be for ordinary users under routine circumstances, while for-pay desktop apps may become even more bloated and specialized for high-end users.

Publication and knowledge. Also in the July+August interactions Aaron Marcus asks "Where do we turn for advice?" and goes on to talk about the user interface tomes. He begins by talking about the third edition of The Handbook of Human Factors and Ergonomics and says

There are two other competing handbooks with almost identical titles: The Handbook of Human-Computer Interaction,by Helander, Landauer and Prabhu, whose second edition published with Elsevier, with 1,582 pages, appeared in 1997, and the equally monolithic The Human-Computer Interaction Handbook by Jacko and Sears, whose first edition of 1,277 pages appeared in 2003... The second of these three handbooks features an editorial board of 14 and 104 contibutors to 62 chapters. The third compendium features 23 advisory board members and 121 contributers to 64 chapters... Other than giving a large number of people opportunities to expound on topics of their expertise, experience, and interest, are these documents worth their weight, to say nothing of their cost?

He goes on to talk about online resources, speculating that "the younger generations may grow increasingly fond of Internet-based resources and eschew the classic paper-oriented resources".

Finally, in the same issue of interactions, Jonathan Grudin has an article you'd never read based on the title: "Death of a Sugar Daddy: The Mystery of the AFIPS Orphans". It's actually a story of looking for the owner of the copyright of classic computer science works. But for me, the most interesting part was his characterization of the state of computer science publishing:

In many fields, journals rely on peer review and conferences use more inclusive approaches as a way to build community. But much of US computer science has shifted its quality showcase from journals to highly selective peer-reviewed conferences. Journal peer review is an awesome resource-free consulting by experts - and journal acceptance rates are higher than those of our conferences, because journals make heavy use of "revise and resubmit" decisions. Last week I spoke to a researcher who said "when I was coming up for tenure I stopped submitting to conferences and just submitted to journals. I've never had a journal submission rejected, but about half of my conference papers are rejected."

OK, I said two things but there is one more. In the July issue of CACM Michael Cusumano writes "What road ahead for Microsoft and Windows?" It's a perfect story for a software engineering class: 50 million LOC and "gridlock".

Making even small changes in one part of the product led to unpredictable and destabilizing consequences in other parts since most of the components were tied together in complex and unpredictable ways. Even 4,000 or so software developers and an equivalent number of software testers was not enough to get Longhorn working.