I would think that a big problem here is to only allow correcting typos, not changing meaning of a tweet - which would suck for a public, fast-paced platform like twitter.
This isn't as simple as counting the number of changed characters, especially when you have to consider at least dozens, better hundreds, of languages.
But then again you can already cram about ten times as much meaning into a single tweet if you write in Chinese or Japanese, so maybe they just don't care.
> I would think that a big problem here is to only allow correcting typos, not changing meaning of a tweet - which would suck for a public, fast-paced platform like twitter.
I feel like GitHub already solved that problem by showing an edit history on comments. Twitter could do the same thing.
Even easier: don't actually publish tweets for 60 seconds, during which time they can be edited.
Facebooks edit history is also hidden in grey text next to the timestamp. I bet putting a "This tweet has been edited" notice actually inline with the content would be effective.
One place I used to frequent solved this by disallowing editing of the original content but allowing the author to add a small annotation after the fact. Doesn't really fix the "correcting typos directly" issue but does allow for clarifications, etc., without the possibility of changing the original meaning that people may have RT/QT/etc.
(You could also extend this to allowing the annotation to be added to a run of text and shove it in the entities object for extra clarity / highlighting.)
Having a time where it's not actually posted would work fine but probably not be super effective: after all, you already have an unlimited amount of time to look at the message sitting on your screen before you actually send it. It's just often very hard to read your own writing for typos. Still it would be a step forward. You could even make it opt-in.
Haha, I swear every time I wish I could edit a tweet it's within 5 minutes after I post it. To be fair, at least with my incredibly limited followers, that's soon enough I can just delete it and post it again without losing too much engagement, but still. I'm just not patient enough to proof-read properly, apparently.
Instead of rendering it as hidden or even as an edit history, render it with an "inline diff" (using the kind of algorithm you see Wikipedia do, or GitHub within a line) as a cross-out of the old content and the next content next to it... changing a lot of text like that would be extremely noisy and obvious--to the point where it wouldn't accomplish any evil goal and would also just be generally discouraged by its nature--and yet would fully solve the vast majority of cases that we want to satisfy.
I don't think it needs to be that in-your-face. I believe an inline "this tweet has been edited" link to a revision history or diff would thwart most abuse.
This isn't as simple as counting the number of changed characters, especially when you have to consider at least dozens, better hundreds, of languages.
But then again you can already cram about ten times as much meaning into a single tweet if you write in Chinese or Japanese, so maybe they just don't care.