Got an interesting email today and have decided to post my reply here as well, as I get this question a lot. I’ve expanded and linked the email out.

To the first point, it is certainly possible to draft a patent license that requests access to products that ensue from a patent. This is done in the “closed” context all the time, and is known as a “reach-through” - it’s also regularly included in aggressive materials transfer agreements (MTAs). It is a fairly standard way to extract rents from downstream products. Here’s a nice analysis of the ways that reach-throughs get examined.
We spent several years on both MTAs and patent licenses at Science Commons, and eventually reached the conclusion that reach-through terms were not something that would work particularly well for “open” patent licenses, whether in attempting to simply retain rights to improvements to the patented technology (which would appear to be the simple case) or to products ensuing from the patented technology. The reasons have little to do with whether or not one could draft such a license, but with whether or not such a license would actually be effective in real life.
There are several reasons for this.
The first is that lots of patents bear on most products, from lots of sources. It is almost never that case that one technology leads directly to one product. When there is a lot of money on the line, everyone is happy to negotiate, and it is not unusual to see anywhere from 15-20% of total revenues paid back in license fees in biotech products. But if one of those patents were to include a patent-left term, the odds are strong that it would never be part of the process or licensed, because it removes the economic incentive (or perceived economic incentive as the case may be) to include the technology in downstream product creation.
We have an example of patent-left in action in the BIOS patent license created at CAMBIA for agricultural biotechnology. It was a set of fundamental gene transfer technologies, patented in order to be made open, and released under a license that granted licensees the right to make and patent products without prejudice or reach through, and only asked that if licensees patented improvements to the underlying technology that those improvements be made available back to all the other licensees. While many signed the license to use the technology, no improvements were licensed back. And that’s not even *product* reach through, where the real money is.
PIPRA is an example of something that is working in biotechnology, which is a pool that allocates resources and works to create greater access through transaction cost reduction rather than through the creation of reach-through or patent-left terms.
Another reason is that patents are not *enabling* rights in the same way that copyrights are. Although copyrights grant the right to prevent others from making copies, the creator who owns the rights owns all the rights needed to license the rights fully to another person, including the right to require them to share alike (copyleft). But in a patent context, especially in a medicines context, the odds are that the right to produce a product is dependent on dozens of patents and materials transfer agreements.
This means that even if you could get a patent on a fundamental technology and convince the owner to license it openly, the production of the product can be blocked by someone outside the “open” transactional circle. Think in terms of chairs. I have a patent on chairs generally, and four-legged chairs specifically. You patent a new kind of chair - a three legged one - and license it openly under patent-left. Your patent keeps me from making a three legged chair unless I sign your license, but my patent lets me keep you, and anyone who licenses your patent, from making any chairs at all unless you license my patent on my terms. (this example is owed to Richard Jefferson from CAMBIA, who has forgotten more about open patents than most will ever know).
The third reason is basic economics, which is that unlike copyrights, patents cost a lot to acquire and thus tend to be acquired by those with little philosophical or economic incentive to make those patents available openly. Copyrights are free and instant, and the advent of digital networks and cheap PCs and cheap digital cameras mean that it’s fast and easy to share stuff, and that we don’t need a large percentage of total content creators sharing in order to have a vast commons. Patents are far more scarce, and far more pricey, and far less likely to be individually owned.
All of this taken together is why we worked on systems that reduced transaction cost, increased transparency, facilitated e-commerce-like effects, and in general encouraged people to think about patent rights as something that they should license non exclusively, but did not attempt to include reach-throughs or patent-left.

The folks at CAMBIA have a new project called the Initiative for Open Innovation which is one of the more cutting edge projects in the open patent space. The idea here is that being able to see the forest of patents is job one, moreso than licensing. Their Patent Lens project is going to change a lot of things. Go look around. It’s worth the time.