Detailed testing and trace analysis has revealed that there is a slowness problem when a NT40 workstation browses to a Netware server.
The following steps are taken in this process:
1 - Standard name resolution occurs through the Netware redirector. This is a bindery lookup.
2 - When a connection is made, NT then tries to open a Named Pipe connection to the Netware server. The process is done by opening a volume named PIPE. If the PIPE volume is located, a file named SRVSVC is located.
3 - The network name for the server is then attempted to be resolved through the Workstation service (Microsoft network redirector). This is done even if the previous step was unsuccessful.
4 - Next, the volumes on the Netware server are enumerated by NT using standard NCP commands.
Microsoft verified that the MUP.SYS update spoken of in the MS article Q171386 does not change what is happening here. This update only applies when a UNC is being used to reference a specific resource, such as a file.
By changing the network provider order in network control panel, the only change made is the order in which the network name is attempted to be resolved through NT.
Microsoft has been contacted and a representative has explained that this has been noted as a bug with NT, but will probably not be fixed before the next release of Windows NT.
A Novell customer found a way to minimize the slowness by doing the following:
Change the NT Netbios node type from H node to P node. This eliminates the Netbios broadcasts and can reduce the lookup time by as much as 7 seconds.
Uncheck "Enable DNS for Windows Resolution". Checking this box could increase lookup times by as much as 15 seconds.
Reducing the number of WINS entries from 2 to 1 reduces the lookup time even further by about 3 seconds.
The customer reported that before making this change, the Netware lookup process can take as long as 23 seconds. After applying the change, it was reported that the time was reduced to 3 seconds.