We had a couple of starting premises when we set out to work on this problem.

First, for more than eighty percent of people, text is more than eighty percent of their information. To paraphrase the famous bank robber who said he robbed banks because that is where the money was, text is where the information is, if for no other reason than that it can be searched.

Text is also long-lasting. Whatever software you use now or used to use, chances are good it can save your data in plain text files. Whatever formats are used in the future, chances are excellent that they will be able to read plain text. We decided that you should always be able to import and export your unformation as plain text. This means you can preserve your information independent of our format and be able use it in other applications.

Second, disk storage is getting bigger and cheaper. Two megabytes of storage once cost half a year's pay and it took up the space of half a refrigerator. We bought a 120 GB drive for a fellow in the office who once worked on one of those machines. He held it in his hand and could only say, "thirty thousand refrigerators!"

This made us decide to store only text and to keep it forever. It not only makes for a very stable system, but also one that uses disk space very efficiently.

When we started out we figured that, like everyone else, we would have our prejudices. We decided to make them work for us by facing up to them and then throwing away the ones that could not be justified by our own experience. These are what remain:

• Old information is inherently valuable. It is more valuable if you have forgotten it. It is less valuable if you cannot find it.
• Information is far more valuable if you know when it was created. Seeing it ordered chronologically reveals how decisions were arrived at or mistakes made.
• Taking care of just your information is not enough. You need to store information about your information (the technical term for this is "meta-information") like a title and time stamp as well. Moreover, we felt we could not assume that the meta-information we thought would suffice actually would be enough for you. There has to be a chunk of freeform meta-information where you can store whatever comments you care to make.
• Hierarchies do not handle information well.
• GREP (all those "ANDs" and odd character you have to key when searching) was created when you controlled your computer by typing. Using the controls of the Mac's graphical user interface is better today.
• Artificial intelligence is usually stupid. It takes control of the situation away from you and gives you one answer and one answer only. Usually it is not the one you want. You understand your information and priorities better than we ever will.
• File systems cannot be trusted. An application that relies on them to organize storage finds that the user moved or deleted something when the application was not running.
• Fashionable formats change and die. The same day someone wanted us to add XML someone else wanted to know if Boswell could read his old WordStar files. ASCII is forever.
• Databases are too fragile.They start out fine, but bog down when you give them huge amounts of data. Because they can so easily be modified, they are far more likely to become corrupted.
• Old ideas that have proven to work in the real world are worth stealing. Hence we use physical metaphors like libraries, archives, and notebooks.

We also had our desires.

• Store a lifetime's worth of text now that the hard drives can handle it.
• Provide thousands of categories, more than an ordinary person will ever need.
• Cross reference the information automatically so time is not eaten up "filing things away" or manually linking every item to its categories.
• Whatever decisions Boswell makes for you, you should be able to see what they are going to be beforehand and override them if you want.
• Always to be able to export as plain text so that the information can be accessed in the future whatever happens. As good as Boswell is, give people the ability to be independent of it if they ever want to be.
• Allow sorting by multiple values rather than just one.
• Handle multiple versions of writings
• "I always wanted a minion."
• "This would have given me straight A's if I had it in school."
• Boswell has be able to satisfy the query, "get me all the e-mails I exchanged with Fred, but not Joe, last April about the Harris project that contain the word 'deadline' and show me the results sorted by time" because that is the sort of request people form in their head before they translate it into whatever their software does.

And we had our hopes -- or maybe they were fears.

• No crashing.
• No data loss.
• Scales up and does not slow down as more information is added.
• Easy to back up.
• Store the text only once.

To keep our focus, we wanted to put some faces on the folks who would wind up using Boswell. We decided to concentrate on students, writers, researchers, and pack rats. Others would certainly find it valuable too, but these defined the boundaries of the territory we wanted to cover.

One of us knows a fellow who writes film maker biographies and conducts interviews with folks about all the many films they worked on and the people they worked with. When it comes time to write about a particular person or film, all those words have to to searched through and the results organized. We kept his problems in mind when creating Boswell because we figured that if we could satisfy his needs, we could satisfy most anybody's.

All of these are what lit our fire. We came up with them before we started designing and kept them in mind while we wrote.