I'm in justusranvier's camp on this one. Not raising blocksize, *at least* roughly allowing for Moore's Law, seems insane, and would just drive some alt to gain significant marketshare.
Bandwidth doesn't increase according to Moore's.
At best, Neilson's
http://www.nngroup.com/articles/law-of-bandwidth/Your source says "10% less than Moore's law." That's not really much difference. Still a lot and still exponential, for now.
Right, That is just the first of many criticisms I have of the current proposals.
I don't like extrapolations used for predictive purposes.
They assume too much.
I'd prefer a method of rightsizing max block size through use of the data of the future which may be derived from the block chain through sums of fees or block size.
This can be done much in the same way that difficulty is generated.
The risks of doing it in this better way may be that it could create some perverse incentives for folks that wanted to "game" the metric.
I'd like to see a proposal that doesn't come with what amounts to a guarantee that it will be wrong because it relies on a guess made years ago.