At Oslo University College, the Netmap displays some layer 2 links as layer 3 links.
The initial view shows all GWs and GSWs and the layer 3 links between them. When adding the SW category to the filter list, but not checking the layer 2 filter, a bunch of SWs appear. This should not be possible, as switch topology comes from the layer 2 topology information (the netbox_to/swport_to fields from the swport and gwport tables).
Layer 3 topology is only derived from the link, core (and external links via elink) nettypes.
Pulling the raw GraphML from Netmap's backend shows the data from the backend is to blame. It contains edges between the routers and switches and mark these as nettype=link and also lists a netident from some link network. The SQL statement used to produce the GraphML also appears to make this mistake.
An SQL output from Oslo University College has been mailed privately to Kristian.
This is hopefully fixed in the new sql-functions. Have to test though. Applet does not honor the new layer-param yet, but that is a task for tomorrow :-)