Until comparatively recently visualisation of geographical information tended to mean one thing. ArcView.
When Google Earth (and the perhaps under rated and certainly under used NASA Worldwind) appeared on the scene visualisation of free geographical information became popular and ubiquitous. Three dimensional spinning globes revolutionized the way none GIS professionals could work with geographical data. At around the same time some comparatively modest open source projects turned into mature easily used products that could be applied to two dimensional desktop map composition.
At the time of writing the best known of these are QGis (Quantum GIS) and Udig. Both are cross platform. Udig uses Java and runs well under Windows. Qgis still performs rather better under Linux, at least in my experience.
Neither of these programs are true geographical information systems, in the sense that they do not of themselves provide very much in the way of analytical tools. They simply allow geographical information to be visualised in a convenient way. Visualisation is a fundamental element of geographical analysis so it should be done well. These open source programs are acceptable alternatives to the traditional ESRI benchmark.
Users of Arcview commonly activate a range of plugins to give the program more features. Qgis provides a connection with GRASS to provide analytical capability, although I would assume that few GRASS users routinely run GRASS through Qgis. Both Qgis and Udig have wms (world map server) connectivity, which means that they can be used to customize the sort of maps that are found on web sites showing wms data.
An interesting recent development in GIS has been the rapid growth in the use of spatially aware data bases. Most major databases (MySQL, Oracle, Postgresql) can now be given a spatial component. However in the open source context “spatial database” has become almost synonymous with the postgis component of postgresql. It is the built in connectivity with posttgis feature that gives qgis and udig their rapidly expanding potential value. Spatial databases allow sophisticated extraction and combination of geographical data. The results usually usually have to be shown as maps. Command line sql statements are not visual of themselves. QGis and Udig can thus potentially provide the full functionality of an Arc* type environment through levering the capabilities of postgis.
At present this very powerful framework is still in a rather incipient phase, in the sense that there are no GUIs that carry out analyses comparable with Arcview/ArcGis. Most of the functionality of the expensive commercial software is now available through the open source alternatives. However while the documentation is comprehensive it does usually assume prior knowledge of data base programming. Most of the current postgis users are knowledgeable data base programmers with considerable experience using spatial sql queries in the context of online mapserver based web sites. However this is changing fast. The WKB (Well Known Binary) representation of geographical information used in postgis databases looks like a good standard and will undoubtedly be more widely used. This should result in the very rapid development of higher level forms of manipulating information in this format.
One advantage of using an online data base as opposed to separate shapefiles is that the coordinate system of all data has to be consistent, or made so through reprojection when the data base is set up.This is especially important in Mexico where data is often provided without details of the projection and datum used (NAD27 and WGS84 are both used and projection details are rarely provided). Also details of how to connect to the data base using Qgis or Udig can be provided in an email, while it can be difficult to send large shapefiles. Both Qgis and Udig allow POSTGIS data to be saved locally as shapefiles.