----------------------------Original message---------------------------- Here at the University of Connecticut's Homer Babbidge Library I've taken a similar approach to the one Johnnie outlined. I've been using a 486DX EISA computer with 2 676 Mb harddrives (total 1.2 Gb) and 16 Mb of RAM as a fileserver. I'm using Novell 3.1.1 with a ten user license. Cost for that... $8,000 for machine, $1,500 for Novell The similarity seems to be suspended there, however. There is no Computer Lab in the Map Library. Rather, I am putting my efforts toward connecting various Labs to the server. I have two public terminals in the map library. One terminal in the Homer Babbidge Library's Reference Area is designated as a connect, but in fact more are capable of connecting. The Cart Lab in Geography is connected as are Geography faculty's machines. In fact the most aggresive use is coming from the School of Business where faculty and lab are accessing the data. This "empowerment" of the user is kind of sneaky because it puts the burden of hardware purchase and upgrade on the user, not the Library. It means that as use of the system grows the library only (!) has to pay for the upgrade of Novell users. Lessons learned, and this is sort of a answer to Mary Ann. When you deliver the data, is that enough? I think no, so I have licensed some software. Ten seats of MapInfo DOS, two seats of MapInfo Win and ten of ArcView. (Also ten of AutoRoute USA and Europe and ten of Map Expert.) I have the login scripts set to applications, not people. The user finds the appropriate data through the OPAC, locates it on the fileserver and down loads a ZIPPED data set with documentation to their machine. Data is stored on the fileserver using an adapted LC class subdirectory scheme. Why MapInfo for DOS? I think the most library intensive use of this spatial data is going to come from the application people, not the geographers (Discussion here!) and most of those people are working on older machines. I understand the limitations of MapInfo for DOS as a GIS, but in terms of geocoding and desk top mapping (which is where many are) it fits the bill. I am putting TIGER data into MapInfo DOS, MapInfo WIN, ArcInfo DOS and ArcInfo E00 formats. The GIS experts have access, put usually don't need the instruction... others do. So I am delivering the data and software to run it on. The server is underpowered. I knew it when I got it, but it HAD to be DOS due to political constraints, and I had only so much money. Maybe it was a bit of a bait and switch... ;-> but I feel strongly that this stuff has to get out on the NET, not be lock in a chronically understaffed map library. So now I am asking for the money to upgrade the system to a Pentium running at 66 Mhz and three 1Gb drives. Data is not only map, of course, but also census. I hope we will be able to provide some sort of access to that data other than spatially. I have converted 1970 and 1980 census data to DBF format on the Tract and MCD levels so the data is importable to most any numeric program. But that is a job for the social science folks. The Map Expert will be on the Reference CD\ROM LAN. Through the magic of scripting, the two servers will be virtually attached to each other. I was not able to afford a tower, and I kind of like getting this stuff distributed. Hope this doesn't muddy the waters any more than they already are. Patrick McGlamery