Hi Cleber,
MST needs to ensure that there is a consistent vlan to instance mapping, else there would be a possibility of bridging loops (see example in the end). That's why the concept of region was introduced: a region is basically a group of switches with same vlan to instance mapping.
As a result, if you are changing the vlan to instance mapping on one switch, MST needs to reconverge because this switch just moved to a different region. The simple solution that was chosen in order to implement that was to restart the MST process from scratch as soon as you change the configuration. There might have been complex optimization, but considering that MST needs to reconverge anyway, they were not going to be very useful. This is basically an issue of MST, not of the implementation. The best work around I know of is to pre-provision vlans and instances in the MST configuration so that they are available when you need them.
Regards,
Francois
Loop example with inconsistent vlan to instance mapping:
Suppose you have two bridge A and B running two instances 1 an 2. A has two ports p1 and p2 connecting to two ports p1 and p2 on B. Instance 1 is blocking p1 on A, instance 2 is blocking p2 on B. Now suppose that vlan X is mapped to instance 2 on A but mapped to instance 1 on B -> vlan X has a permanent loop between A and B. MST prevents this issue by putting A and B in different region, and forcing all the vlans to be handle by a single instance (the CIST) between region.
