Practical Internet Groupware

Shared Versus Annotated Data Stores

At its most minimal, groupware is just software that can read and write a shared data store. But wait a minute; doesn’t a simple Novell NetWare file-sharing network qualify as groupware according to this definition? Yes, it does. Who among us hasn’t played out the following dialogue?

You: “Where’s the schedule?”

Me: “It’s on drive T:, july98.xls.”

More of us do more of our business this way than we care to admit. We realize that it’s a flawed mechanism, but we tend not to analyze why it’s flawed. There are two reasons. First, the file-sharing approach relies on an unannotated data store. A tree of filenames can describe only so much about its contents. Things got much better when long filenames became standard, but a bare hierarchical namespace remains an impoverished way to describe a data store that houses the intellectual capital of an enterprise.

The second flaw with the file-sharing approach is that it isn’t, in and of itself, a mode of communication. After you copy july98.xls to drive T:, you have to tell someone—maybe everyone—that it’s there. Hollering “It’s on drive T:” over your cubicle wall to everyone within earshot was the time-honored way. But now that everyone is either working at home, or traveling, or in a satellite office on the other side of the continent, we need to project our voices through digital networks. Enter email.

Email appears to solve both of the problems with file-sharing. It does annotate the data store, in the sense that messages can richly describe and comment on attachments. And it is, obviously, a mode of communication. But do you see what’s been lost? We defined groupware as software that can read and write a shared data store. Email doesn’t do that; typically it reads and writes personal data stores.

Email can be used to annotate file systems, and in fact it tends to supersede them. We like to use mailboxes to manage documents, because there the documents are surrounded by useful context. We know who wrote them (usually not obvious in a raw file system), when (file systems do know this too), and most crucially, for whom and why. The name july98.xls doesn’t distinguish between “the version that includes Bob’s changes, as requested by Sally, but uses last week’s numbers” and “the nearly final version, subject to formatting tweaks by Richard and financial review by Ellen.” We’re addicted to email, and rightly so, because it helps us relate business documents to business processes.