BOT2001 Report
Where Can I Put My Bot?

By Brian Proffitt

Sometimes it's the simple things that can throw you completely off track. You can have everything you think you need to build the world's coolest bot: the programming knowledge, the flow of the application, the technology--you would think you were all set, right?

Then it hits you: where are you going to put the bot? Will it be on the client computer? Or the server?

Before you start thinking this is a cut-and-dried situation, you might want to hear what Ken Shapiro, Chief Technologist for WorldFlash Software, Inc. has to say.

Shapiro spoke at this year's BOT2001 conference, comparing server-side versus client-side intelligent agents for the audience. He also spent some time examining the various privacy and ethical issues that might arise depending on the positioning of a bot.

Bots which load on the client's computer offer some distinct advantages over server-based bots, Shapiro explained, but only in certain areas. In other areas of concern, the exact opposite is true.

For instance, a client-based agent will utilize the user's CPU resources, not those of an expensive server farm. Client bots can usually obtain more information about the user, such as from the client's hard drive. And, because the most of the bot is on the client, bandwidth is usually not an issue.

There are some disadvantages for client-side bots, though. There is typically at least one major download involved, and bot developers will have to have their bot ready for installation on any number of different configurations.

Configuration is a much simpler process for server-side bots, since the bot will be installed on just one type of machine, typically. Since the bot never has to be placed on the client PC, then the large download process avoided too.

However, servers can be expensive to manage and maintain, and depending on the nature of your bot, you may have to run a lot of servers. Also, the traffic between the bot and the clients will mean you will have to carefully manage the bandwidth from the servers to the Internet. Finally, only the information about the user that's available on the Internet can be used.

Privacy, often a major concern for end-users, will have to be managed differently--depending on how the bot is positioned.

A server-side bot's databases, for instance, will need to be carefully protected, since a great deal of personal information will be stored in them. That much information in one place could spell disaster if not properly protected.

For the client-side bot, care must be taken not to violate the end-user's privacy by back-channel monitoring of the user's behavior with their PC, such as where and when they surf the Internet.

Shapiro presented three privacy options to give to users no matter which kind of bot is used. First, he said, give the customer complete disclosure on what kind of information your bot will collect and where it will send that information. Next, give the user a chance to opt out of sharing personal information. And finally, give the user the chance to modify their personal information at any time.

Using these techniques will give the user enormous peace of mind, Shapiro stated.