There are a lot of theories, and a lot of systems. The bottom line is, you have to find something that works for you, whether it's an established, respected system or something you put together yourself after observing how your body reacts to different types of training.

I've enjoyed the perenniel long run distance discussion. I am in the camp that says the Jack Daniels 2.5 hour limit is excessively restrictive; but there are arguments to be made in its favor. This training cycle, I peaked at a 22 mile long run, which was ~32% of my 68 mile peak week. I did not honor Jack's 20 mile limit or his 2.5 hour limit; but I came in under a 3 hour limit, as some other systems advocate. I ran my last 22 mile with a negative split, running the last 8 miles faster than the first 14. At the end, I felt like I could keep going for another 4 miles. This is in contrast to my preparation (such as it was) for my first marathon, where a 22 mile run capped a peak week of 41 miles and I felt like maybe I could keep running another 2 miles at the same pace. (And I didn't bring that one in under 3 hours.)

From other runners, I hear different stories of what works. I know a guy who trains to 26 mile long runs. I asked him about it. He said he trained to 20 miles for his first marathon, and that wasn't enough to prepare him for the marathon distance. So he found a training program online that went up to 26 miles, and had better marathon experiences. And he turned in a double BQ at Rochester 2015 (in the window to qualify for Boston 2016 and 2017), on a new, challenging course. For reference, a very good local runner ran Boston 2015 in 3:00, but took 3:11 to run Rochester 2015 and said the course was challenging.

I've also talked to a marathon runner who told me there's nothing quite like a good 4 hour training run. Not everyone believes even in a 3 hour limit.

On the other end of the spectrum, my coach came back from Achilles tendinitis in 2015, spent a lot of weeks not running a step, and only worked up to a 14 mile long run before turning in a BQ performance at Niagara Mohawk, while weighing 30 pounds more than he wants to for running efficiency. But to be fair, he has over 5 decades of running and coaching experience and knew exactly what he was trying to do. The training plan he wrote for Boston has 20-22 miles for the longest run, three times; I hit 22 on that plan twice this cycle.

I agree that a strong base is the most important thing for completing the marathon in good shape. I suspect that the measurements of long run as X% of weekly mileage aren't really measuring the right thing. If I run 60+ miles in a week, I'm forced to run over 10 miles more than one day. That can be done with easy runs that look long, or with doubles. @Stoshew71 has said he runs doubles because that fits his schedule; I've run 10-12 mile easy runs because it's easier on my schedule to run longer in one chunk than to accept double the time overhead of pre-run and post-run routine. Those longish easy runs are doing something for me that I didn't get when I only did true long runs and had less weekly mileage supporting them.

Having said all that, I also suspect that optimizing training for a marathon is different at different paces. I ran 22 miles in just under 3 hours. At a 9 minute mile, 20 miles would take 3 full hours. At a 12 minute mile, 20 miles would take 4 full hours. Physiological effects aside, there's a big difference in scheduling requirements to get the miles in at different paces.

And a quote to think about from my coach. We were talking about preparing for a 100 mile race. He said, "Obviously you can't get any 80 mile training runs. But you can do two 40 mile runs back to back on Saturday and Sunday."

There are ultra runners who do not limit themselves to 26 mile training runs. But 50K and 50 mile races are not run as fast as marathon pace.

I believe many of the restrictions we hear touted spring from an assumed goal not of being able to run 26.2 miles, but of being able to run 26.2 miles as fast as possible. That's an important distinction. There's a lot more wiggle room when training for completion than when training for an aggressive time goal.

@MobyCarp Jack Daniels doesn't specifically say that you shouldn't do 20 milers. He advocates either 25% of your weekly miles or 2-1/2 hour limit (whichever you hit first). So he has no problems with you running a 20 miler as long as you are running at least 80 miles a week and can run that 20 miler within 150 minutes.

But point taken that it is very restrictive. But he also is very conservative in all his training methodologies.

I mean, there are training plans that call for a max of 16 miles, which is only ~62% of total race distance. So yes, that's totally possible to get a good base with 14 miles a few times a week. Though that would be a heck of a time commitment!

For a marathon, I think those 18-20 mile runs are key for psychological readiness but not necessarily physical preparedness. A friend of mine just got to the taper period of her first marathon training cycle, and she said that doing 20 miles was a huge confidence boost (and also reality check), but she didn't up her mileage the way I encouraged her to and she was feeling it. When we did 22 last week, her total miles for the week was just 39, so her long run was 55-60% of her weekly miles, and she was saying this morning she's still sore from Saturday. So I think it has less to do with "longest run" miles but cumulative mileage.

TL;DR: So could you run a marathon on base building alone? Sure, if you wanted to run 14-16 miles multiple times a week! I think that the cumulative mileage that way would get you to the finish line, for sure.