There is nothing wishy-washy, slippery, or shoehorn about the proposal. It is very consistent with the thoughtful analysis that helped us arrive at our current subnet topology.
DFINITY conducted a very thorough analysis of the existing decentralization characteristics of the node distribution for the purpose of assessing how strict the decentralization targets should be set. At the time, when all subnets were global in nature, it made sense to set the data center provider (owner) target at a subnet limit of 1. Now that we have a need for regional subnets, the same thoughtful analysis is being applied. In the US we only have 10 unique data center owners. Hence, the subnet limit of 1 doesnāt make sense and is not practical. Given that these data center owners are large corporations that enter into significant and legally binding contracts with node providers that establish clear expectations for privacy and security of the node machines that they host, there is no logical reason not to relax the decentralization target for data center owners to 2. This is no different than the analysis and conclusion that was reached on the decentralization target for the country subnet limit, which is currently set at 2 for smaller subnets and 3 for larger subnets.
We need to be making sensible decisions about subnet topology, which is what this proposal achieves.
My point is that if the NNS is prepared to weaken the IC Target Topology here and there (rather than maintain, or strengthen it over time), then itās anyoneās guess what decentralisation safeguards weāll have in several years time (with more demand). I donāt think thatās good at all, and itās why Iād like to see a more rigorously specified topology which we can depend on. Demand, and relative availability of nodes, is not static with respect to any subnet.
Iād support a proposal that redefines the IC Target Topology as a set of tolerances that apply to all subnets, with special cases only when they make sense for the right reasons (as is currently the case as you mention), i.e.
Larger subnets can manage slightly more forgiving tolerances on less critical dimensions
Subnets with specific roles where the role directly relates to a dimension (e.g. country and regulatory compliance)
If some dimensions really are less important in general, the standard tolerances can express that.
This would be a good opportunity to represent Continent, and possibly ISP, as additional dimensions.