Right now, with BIP125 quite widely adopted amongst wallets and services, it is quite safe to assume that a transaction signaling against replacement will not be replaced.
Not really. If this were true you would have exchanges accepting deposits on zeroconf. Does that happen today? Nope -- almost nowhere. And the ones that did so (e.g., that one bitcoin ATM operator), got thumped.
Now for certain transactions (e.g., low-value, retail transactions) or where the sender can be trusted, sure -- zeroconf can be acceptable because the rare instance where loss occurs is offset by the profits from the vast majority of transactions where no loss occurred.
So, ... I'm fine with bitcoin where Full-RBF exists. Especially now that we have Lightning network for fast transactions.
reply
You're right! I should have been more cautious in my formulation. I actually had the in-person everyday purchase use case in mind (eg buying groceries with Bitcoin), where you either rely on 0-conf or embrace Lightning ; and somehow "forgot" that it is far from being the general case. I just pushed an edit to the article, thank you for comment!
reply