1) For load balancing could be used 2 routes having different administrative distances ? Or for load balancing always the routes must have the same administrative distance?
2) Here's my understanding of the impact the "no ip classless" command has on router behaviour:
"no ip classless" command has no effect on router behaviour when supernets and default route are learned via OSPF or IS-IS, but has effect when defining a static default route and also when supernets and the default route are learned via RIP, RIPv2, IGRP, EIGRP or BGP.
First please confirm if my understanding is correct and second, if the case, what is the explanation for this different behavior, dependent on the routing protocol type?
10.1.1.0/24 |-R1-| 10.1.2.0/24 |-R2-| 10.1.3.0/24
If a static default route is configured in R1, then ping between PCs on 10.1.1.0/24 and 10.1.3.0/24 will not work, because R1 will drop the packets.
If default route is learned via OSPF or IS-IS then ping will work.
1) EIGRP and older IGRP can perform weighted load balancing using the variance concept:
the second route has to satisfy feasibility condition or it cannot be considered.
If metric of second path is 2 times the feasible distance this link will be used 1/3 and best link 2/3 (for this is weighted).
I'm sorry I haven't one explanation for this.
We can guess that IS-IS being not IP based can behave differently.
Probably it comes from link state properties:
these protocols are inherently classless they haven't an auto-summary at major network boundary and this cannot be enabled : a topological area border is needed to perform summarization.
From this can come the different behaviour.
I agree the behaviour is evident when even if a default route is available: the classful routing device will discard packets for an unknown subnet of a connected major network even if a default route or supernet is available.
Hope to help