Mark, thank you for your extremely lucid and well-informed presentation of the pros and cons of current implementations of Web servers and clients. I think the issue of "where does the processing occur" over the network or locally, and do you have to pump all these bits through this narrow 1200 baud passageway has *alot* of relevance for museums. Let's imagine two models of providing images and data of the XYZ Garden's 8 million specimen herbarium collection (and I *must* emphasize that this is all personal speculation, though I am on the NY Botanical Garden staff, I have no authority or responsibility for computerization.) The first is to sample and store all the visual data on a server, with text data associated, and all the associated search functions, pointers, etc. This server is at the Garden, on the Web, and accessible to all and sundry with Mosaic or other Web client tools. The second model is that low density images, for example, or just text descriptions of specimens are distributed on CD-ROMS by subscription to libraries and universities (I know...gasp, not universal access, but is that so important for the *many* specialized collections that are out there?). Search tools are either standardized, or provided on the CD-ROM itself. When the search is narrowed to the point where additional information is required, or more detailed images are needed, the user may access the server through the NET or whatever. Dare I suggest that this is a revenue stream to the museum, analagous to a publishing program. I mean its *expensive* to database this stuff. In another example, imagine that there is a standard engine for interpolating video images (actually I assume that there is, but I'm not technically up with it.) This is stored locally as software or firmware or whatever on the client. Then all that would be required is the transmission of "clues" from the Museum to provide a full motion video tour of the museum, or whatever interactive content the museum wanted to offer. All I am saying that Negroponte suggested is that taking advantage of computing power on the client side could reduce the clogging in the pipeline. I mean we all have pretty zippy machines as clients nowadays, and I am pretty sure that the bottlenecks don't take place in our processors, hard drives, or video sub-systems. Eric Siegel [log in to unmask]