Well, hold on. The OP told us back in post #5 that . . .
Whereas your methodology specifically defines “street” as a continous stretch with the same name. So its hard to know how much this tells us about the geography or geometry of a city, and how much it tells us about the city’s conventions or practices with regard to street names.
Which is exactly what Mr Downtown calculated.
Maybe some streets and motorways are cut off at the edge of the metro area, but this should not change the average much.
Sure, this methodology would be problematic in a city where a significant fraction of streets and alleys have no names. So far, though, it seems we are seeing some variation among denser versus less-dense cities, and organic versus planned cities, supporting posters’ intuitions that the former have shorter streets (and probably blocks). If one needed to dig deeper into the geometry, we could further compute the network’s fractal dimension, intersection angle of segments, etc., but I think the gross number, which I asked about in the OP, is somewhat interesting. The average street in Chicago is 5 times as long as one in Paris!
I never asked what it all has to do with numbering practice and naming conventions (if I am not mistaken there have already been threads about those), but feel free to speculate 
I wouldn’t be so confident. Some of those motorways that are cut off at the edge of the metro area will be thousands of kilometers long, and one problem with using the average as a measure of typical value is that a small number of extreme outliers can have a significant effect on the average.
Yes, it is. But - perhaps I misunderstood you - when you posted that “I don’t care about name changes . . . just actual street geometry”, I understood that to mean that if you have a continuous length of built=up thoroughfare that function as a street, different sections of which of which are variously A Street, B Avenue, C Boulevard, D Street, D Street North, E Road and F Street, you would regard that as a single street, because you don’t think name changes are relevant. Whereas Mr Downtown’s methodology would treat it as seven different streets.
Baseline St. in Southern California runs from the base of Mt. San Bernardino to Rancho Cucamonga, a distance of approximately 36 miles. It was, as the name states, the base line for the survey and plotting of most of SoCal. The street runs through Highland, San Bernardino, Rialto, Fontana, Ontario, and Rancho Cucamonga, along with a hunk of unincorporated territory.
Well, since we’re here, that holds true in most of Chicago (as that article points out). From Madison to 31st, it’s 12, 10, and 9 blocks to the mile until it resumes the normal pattern at 31st Street.
^Don’t confuse the house-numbering scheme introduced in 1909 with the subdivision pattern. You’ll notice that out on the West Side, some numbered streets (generally 17th and 29th) are missing. When the streets near the lake were given numbers in 1861, they simply numbered them in sequence as they were, one number for every street, no matter how close it was to the previous street. That’s why the area between 12th and 39th doesn’t observe the modern 800 numbers to a mile rule. Twenty years later, standard practice was to create 330- by 660-foot blocks, or eight to the mile in one direction and 16 to the mile in the other. So as areas developed further west from the lakefront, there were sometimes fewer streets than available numbers and someone—the developer or the Superintendent of Maps & Plats—had to decide which one to omit.
Yes, you’re completely right. I was treating 100 address numbers as a “block.” The actual size of the blocks, though, once you get south of Roosevelt is 1/8 of a mile in most cases (as you say). That said, so far as I can tell, Madison to Roosevelt pretty much all the way to Central has shortened blocks, 12-to-the-mile in most areas. But there’s all sorts of irregularities, depending on where exactly you are in the city.
A highway necessarily has a higher speed limit than a street. About 45/50mph would be the cut-off point.
Note, that only helps if your dataset includes speed limits but not the road classification.
Baker Street in London didn’t have a 221 when Sherlock Holmes would’ve been solving crimes back in the late 19th century, but the street was later renumbered so that tourists and fans would be able to say “This was the spot.”