Mastodon itself as a service and protocol, is pretty wonderful.
But in terms of namespace, domain-name, and instance issues, you've got pretty much all your traditional Internet concerns, with a few new twists added.
My first primary instance has gone the way of the dodo, and its domain is now parked somewhere entirely different (https://mammouth.cafe). Numerous other instances have also come and gone.
My current instance, run initially by an anonymous administrator, was transferred to a Japanese concern, who don't do much to keep anyone advised of service outages or their causes/resolution. This is an instance with 56k registered users, large by Mastodon standards.
Then there are the issues with blocking and federation, which don't immediately concern namespace, but could as blocked instances die, and potentially others occupy their namespace, or they change their identity to evade blocks.
There's been a lot of agitation over the dot-org PIR self-dealing and corruption, as that's a fairly commonly-used TLD, either directly by instances or for related resources. The whole DNS / registry system is a bit of a mess.
The fact that between the Internet and DNS there's no real allowance made for either space or time is becoming an increasing stress point. What solutions might exist isn't clear, but assumptions of the 1970s and 1980s are not being sustained 30-40 years later.
As I've noted a few times recently, a dictionary bibliographic entry typically consists of little more than name, dates, nationality, and profession. Webster's 11 Collegiate includes some 6,000 or so such entries. There are roughly 1 million times more people than that now living, and another 100 or 200x more who have ever lived.
In digital systems, there is no distinction between use and mention -- the name is the thing and the thing is the name. This is convenient in narrow scopes (time, place, scale), and ... exceedingly inconvenient over broad scopes (time, place, scale). UUIDs address the collision problem but not the convenience. Convenient shortcuts address convenience, but not uniqueness requirements.
Since meatspace identity is who you are, and other references (names, nicknames, registration numbers, etc.) are not identities but identifiers, the problems aren't as manifest. Your actual identity cannot be merged with another (though numerous related aspects and relations can be). Online, if A and B are suddenly both called "A", then they are both "A".
I think we'll be sorting this out for a while yet.
But in terms of namespace, domain-name, and instance issues, you've got pretty much all your traditional Internet concerns, with a few new twists added.
My first primary instance has gone the way of the dodo, and its domain is now parked somewhere entirely different (https://mammouth.cafe). Numerous other instances have also come and gone.
My current instance, run initially by an anonymous administrator, was transferred to a Japanese concern, who don't do much to keep anyone advised of service outages or their causes/resolution. This is an instance with 56k registered users, large by Mastodon standards.
Then there are the issues with blocking and federation, which don't immediately concern namespace, but could as blocked instances die, and potentially others occupy their namespace, or they change their identity to evade blocks.
There's been a lot of agitation over the dot-org PIR self-dealing and corruption, as that's a fairly commonly-used TLD, either directly by instances or for related resources. The whole DNS / registry system is a bit of a mess.
The fact that between the Internet and DNS there's no real allowance made for either space or time is becoming an increasing stress point. What solutions might exist isn't clear, but assumptions of the 1970s and 1980s are not being sustained 30-40 years later.
As I've noted a few times recently, a dictionary bibliographic entry typically consists of little more than name, dates, nationality, and profession. Webster's 11 Collegiate includes some 6,000 or so such entries. There are roughly 1 million times more people than that now living, and another 100 or 200x more who have ever lived.
In digital systems, there is no distinction between use and mention -- the name is the thing and the thing is the name. This is convenient in narrow scopes (time, place, scale), and ... exceedingly inconvenient over broad scopes (time, place, scale). UUIDs address the collision problem but not the convenience. Convenient shortcuts address convenience, but not uniqueness requirements.
Since meatspace identity is who you are, and other references (names, nicknames, registration numbers, etc.) are not identities but identifiers, the problems aren't as manifest. Your actual identity cannot be merged with another (though numerous related aspects and relations can be). Online, if A and B are suddenly both called "A", then they are both "A".
I think we'll be sorting this out for a while yet.