Colleagues, I'm not going to repeat the corrections offered by Greg, but having known Bruce since the Melbourne IT investment in NeuLevel circa 2001, that is, from shortly after the selection of the first six entities to become registrars accredited by the corporation several Boards and Executives ago, during the first "6-10 new gTLD" round, I'm going to comment.
I believe it was deliberately set up as public benefit rather than a member organization - to avoid the situation where the members become limited to say gTLD registries and registrars and hence it ends up operating primarily for the benefit of the domain name registration industry.
Bruce errs, as Greg has pointed out, in that California allows both member and non-member forms of non-profit corporations, but he does not err in attributing fundamental importance, even to the form of incorporation, to an attempt to avoid, what is known as "agency capture" in Administrative Law -- and Network Solutions loomed very large in the minds of many involved in the formation of the initial constituencies and in the minds of the founding staffers. Whether "agency capture" was avoided is a matter of opinion, or perspective, but regardless of Bruce's attribution, the intent to {rule make | technically coordinate} independent of the legacy monopoly incumbent, and subsequently its sales channels with market power, was genuine and Bruce correct in expressing general intent of the first and early Boards -- though not to the issue of choice of corporate forms. The sentiment Bruce shared is correct, though the specifics, 16 years later, are not.
Any move away from a public-benefit corporation to a membership corporation - would need to carefully consider how to ensure that the members are reflective of the broader Internet community and don't become limited to a few members as interest in "ICANN" drops over time. I.e. a failure scenario of membership organisation is what happens to the membership base over time and how it can be protected from capture. I have seen some membership based ccTLDs get into problems when their membership becomes dominated by domain name investors for example.
Without commenting on the problems of ccTLDs I will share my personal experience as a member of the Membership Implementation Task Force (Indigenous Peoples with John Afele of Ghana) in 2000, an effort which eventually lead to the notion of a source of input to the corporation outside of the parties, most with some degree of economic self-interest, which formed the first constituencies -- this was the "At Large" project, which eventually was incorporated into the Bylaws in October, 2002. I still consider the "membership problem" non-tractible. One either ends up with a wicked small group of "members" capable of governing a corporation (in the usual sense) which affects directly the lives of many orders of magnitude more people who have no "representation" in the consequences of these governing decisions, or one ends up with a wicked huge group of "members" incapable of governing a corporation (in the usual sense). We would be less than accountable ourselves were we to fail to observe that some of the Bylaws entities or their component sub-entities have, as Bruce observes, "become limited to a few members as interest in 'ICANN' drops over time". I've observed this first hand in the PSO, in parts of the GNSO, and in one of the AC entities. Protecting the integrity of the membership base is necessary to prevent institutional capture. The sentiment Bruce shared is correct, though the specifics, 16 years later, are not, and I broadly share the sentiments he expressed. Cheers, Eric Brunner-Williams Eugene, Oregon