The thing is, by definition, the registry is a monopoly for a certain TLD. So the idea was that those registries need to be regulated. What if they would just ask insane prices?
A registry has an 'easy' part, and that is actually maintaining a list of names and associated data, and providing that as DNS zones.
The hard part is dealing with individual customers. Providing service, collecting money, etc.
So the idea at the time was to have lots of registrars competing on what service they provide for what price and having more or less regulated registries that provide a fixed service at an agreed price.
By and large this system worked.
Early on, it was not clear if large zones like .com would actually scale. So there was some restraint of registering silly names, because that could lead to problems.
At some point, computers got fast enough and big enough memories that large zones were no longer a serious worry and everybody moved in to get all the nice domains.
I'd say the many-registrar-middlepersons approach system worked badly. All we needed was the registry to do very straightforward things, under contract. Instead, we got additional complexity and inefficiency, gobs of money being handed out to opportunists, people newly arriving to the Internet interfacing with certain "registrars" that used dark patterns to upsell, whois frontends rumored (and I've seen this myself) to grab the domain you're checking and then offer to sell it at an inflated price, and typically the domain name you wanted for your startup or other project was already registered by a squatter (I saw this many times) requiring you either to pay up the gouging scalper rates or keep trying to outpace the dictionary combinatorics attacks, and domain names more accessible to the deep-pocketed.
Think about the current state of ICANN. And now ICANN has to award contracts to monopolies for gTLDs like .com and .net.
What could possibly go wrong?
The registry part is easy to specify.
For the registrar, look at how google normally operates: everything they can automate works well and is cheap. As soon as something goes wrong and the they lock you out of an account it becomes hopeless. Imagine that google would be the only registrar for .com or .net
Of course, with a free market come shady parties. But with a bit of effort it is possible to find registrars that work quite well.
With just one monopoly per gTLD, it is likely that bad things will happen in the long run.
I don't know why we're still talking about monopolies. It should be a contracted clerical registry doing a very straightforward thing, according to rules.
For big zones it is relatively cheap (i.e, the cost per domain is low).
Since the invention of anycast routing, it is technically no longer complex.
If a continent fails, then either all BGP announcements are already gone, or you need a way to stop them. Then traffic will be routed to other continents.
It is not trivial, but for big zones if you get 1 euro per domain per year, you can do a lot of stuff.
And that is the sort of prices people cared about back then.
On the other hand, providing support, nobody has any idea how to spec that. What is a reasonable cost? The same goes for payment options.
A registry has an 'easy' part, and that is actually maintaining a list of names and associated data, and providing that as DNS zones.
The hard part is dealing with individual customers. Providing service, collecting money, etc.
So the idea at the time was to have lots of registrars competing on what service they provide for what price and having more or less regulated registries that provide a fixed service at an agreed price.
By and large this system worked.
Early on, it was not clear if large zones like .com would actually scale. So there was some restraint of registering silly names, because that could lead to problems.
At some point, computers got fast enough and big enough memories that large zones were no longer a serious worry and everybody moved in to get all the nice domains.