General agreement with the recommendations; some comments
I won't discuss the general argumens in favor of vertical integration now, because I fully support them. It does promote competition as no other measure, it prevents discrimination and it is critically important in a market where actors, both registries and registrars, are "artifficially" created contractual parties with privileged access to the source (domains) byvirtue of, precisely, their contractual roles (and by artifficial I don't mean anything derogatory, just underline the fact that the model is this way by design, not by nature).
I therefore agree with maintaing the current structural separation.And I've been proposing the hybrid model for a long time. I would only urge ICANN to include the non-ICANN contractual counterparts acting as a registry operator. In sTLD contracts, and in many future gTLDs the registry platform provider is not the (contractual) registry, but a third party. If this third party is also a registrar, it should be also prevented from acting as such in those TLDs. As an example, CORE acts as registry operator for both .museum and .cat, and while being a registrar in all gTLDs, it is not acting as such for those two concrete domains. Because both CORE and those registries so agreed, as ICANN rules in this case are less clear.
Secondly, I find the recommendation on "single-owner TLDs" difficult to implement. Soem of the arguments make little sense (if a company does not want to realse to the public a yet-to-be-announced product, being its own registrar is just one of the many solutions; registries as sTLD demonstrate, have rules not just on who and how can registrer a name, but also waht can be registered. Not allowing "free, any name" registrations would be enogh; more to the point, if it is single owner... who else could register those names to be kept secret? Nobody. And why would them register them, if there is no risk, by absolute legal, technical and procedural inability, that anybody else register those names?). But let's assume it makes sense to exclude registrars. What then is a "single owner TLD?". .BIGCO registering the names for theri products, departments... OK. And its employess. OK... but less so. But what happens if they want to allow their (legally indepemndnt) resellares, retailers, consulting partners to also register under .BIGCO? IS a single one of these domains enough to change the model? Perhaps the problem with "single-owner TLDs" is that they bring more noise to the DNS without really providing much, if any, benefit to the public, common interests at all. But that is a different model.
Finally, a comment on the dismissal of vertical integration on sponsored TLDs. Or we we want, small TLDs in general. The argument is dismissed as a single registrar can perform the registrar function. Sure. Except for the absence of competion. And the main problem with "small" registries is that they attract a small number of registrars, which is logical, but a special kind a registrars: hihg-margin, service intesnive registrars. Which is good. Except that it provies very little flesibility and actual diversification. And seriously limits the ability of registries to promote the TLD. For a small TLD getting sufficient attention even from its accredited registrars, or proposing anything outside the usual, cohesive practices of those high- margin registrars is very difficult, Much more difficult than it would be for a company able to afford and manage a TLD to effectively negotiate with one or more registrars. So perahps ICANN could consider allowing registries managing a TLd under, say, 10.000 names, to act for such time as registrar (NOT as single registrar).
Amadeu Abril i AbrilDisclaimer: I work for CORE Internet Concil of Registrars on new gTLD projects, and was CEO of .cat registry at the time of its launch, so read my comments with those present and past interests in mind