Only bad actors will disobey and they will have to modify their software in order to do so. This doesn't provide absolute protection against bad actors, but since most bad actors don't own a time machine, it reduces the scope of the harm they can enact.
Consider the adversary "angry ex-boyfriend." Let's assume he wasn't always angry and isn't a sociopath. By the time he has become angry, the sensitive posts have already been deleted. This makes a difference in real life to real people.
There is indeed a user-interface concern, to not over-promise to the end-user. But that doesn't justify leaving such a useful thing out at the protocol level.
You mean besides me, of course. XD
EDIT: It's a joke people, for Pete's sake.
I would not put it past Facebook (or other businesses which are addicted to harvesting user data) to behave badly against networks like this. However, for the sake of their reputations they'd still probably do it quietly, which means many of the person-to-person attacks that deletion protects against would still be thwarted.
This is a problem the same way Facebook's privacy controls are a problem. Facebook themselves are not bound by them, but they're still useful if you want to protect your data from other users of the platform.
And FWIW, I think most of the big crawlers respect robots.txt - this is the same sort of thing.
I have an hourly bup backup for each of my servers, that goes to a backup host. I am not a bad actor, have not modified any software, and yet if I run an activityhub server I will have a complete timeline at 1hr granularity, so basically anything not deleted within minutes of posting.
(Considering a switch to borg backup when I have the time to really evaluate it)
Reasons to be skeptical of it: well there's the obvious ones, you might want to hold on to some content coming your way and lots of people are disappointed to see things disappear. Mutating networks have some dangers for sure.
Reasons it to want it, or that it's there: well, this is how all the existing federated social networks work. ActivityPub was designed in the context of standardizing basically something that projects like Mastodon, Diaspora, MediaGoblin, GNU Social, etc can use. And that is how things work now... mutation and federated deletion has been part of the existing federated social web for a decade. And we're trying to build something people can hook into their existing systems. People may also change their minds about things. Should they have the authority to remove their own posts? An open question, but at the very least, whether a server removes the content or not, a Delete is a request to do so across the network.
However, yes, there is an alternative structure one can imagine for these things, and that's something a bit closer to something git-like in the way that branches point at commits, but may change at which commit they point to over time, but the commits themselves never change. That's a desirable thing to explore, and I've blogged about it: http://dustycloud.org/blog/an-even-more-distributed-activity...
And we're even looking at exploring it, possibly, in the Social Web Incubator Community Group: https://www.w3.org/wiki/ActivityPub_extensions#Possible_futu...
BTW, if you're interested in such things, the community group is open to everyone and is hosting weekly meetings https://www.w3.org/community/swicg/ and we'll be hosting a meeting next week https://www.w3.org/wiki/SocialCG/2017-05-31
However, I think one of the big things that cryptonerds and other people deeply involved in developing this type of technology overlook is that most people don't want a system that tells them "good luck retracting anything you ever publish". Integrating this as a "feature" is actually an active barrier to adoption.
Yes, it's true that you can't ever guarantee all copies of something have been destroyed after they've disseminated, but you can still provide the facilities necessary to signal that the content owner wants that to happen, and create an expectation that blessed clients will observe that signal internally.
We need to provide mutability at least on par with the current experience if we expect one of these alternate social protocols to pick up steam. These are popping up like flies recently, and most overlook this.
Additionally, though the technology can't force other instances to delete things, other concerns perhaps can. Consider the potential legal implications of deliberately ignoring delete requests.