Maybe I'm way off.. Maybe its view of KISS but as engineers we should
all be looking for the simplest answer. To me they key in Dragos'
post was usage. All physics aside, the warm weather (seasonal) people
go out more, use the internet less. In cold months, we stay in, use
the net more. As for document any of us that run networks have seen
this well document going back many years in our mrgt graphs. But
then maybe he was refering to the physics, and I just try to simplify
things to much.
Usage has an effect on overcommit, yes.
However, when you notice that the average connect speed goes down for a
day or two after a cold heavy rain, that's not usage. There are not
more than one modems connecting(*) to a port. So you have established
that something about the quality of the physical layer has been affected.
(*) In the late 1990's, I heard the most astonishing claims made by a new
entrant into the Milwaukee ISP market, about how some of the "other" ISP's
"shared" lines between customers and this decreased your speeds. They had
no clue who I was, so I engaged their technical person for a while who set
out to convince me that other ISP's really _did_ do this mythical line-
sharing - multiple modems to one port. Until I started talking about the
technical aspects, that is.