Hacker News Reader

Comment by abledon | Hacker News Reader

Is the 'edit' button feature actually a hard feature to implement in twitter (due to their scaling architecture)? I remember reading a passage in that "Designing Data Intensive Applications" that explained how their architecture was different if the account was say, a celebrity account which needed to push out to a lot of followers, or an account that had few followers and tweeted very little etc..

View on HN

They probably made assumptions in their architecture that tweets can never be updated, and that they can only be created or deleted. Depending on how ingrained this assumption is, it could be quite difficult to change. Caching, replication, eventual consistency - lots of hard stuff. Until relatively recently, I believe AWS S3 didn't have read-after-write consistency for updates to existing objects - I doubt that was easy for them to do.

I've seen lots of bad hot takes from Twitter users who have no idea how large systems like this are designed. Dumb stuff like "just do a SQL update" or "use React".

There's all the political decisions to make as to what do you do with existing Likes, Retweets, etc, what oversight will there be, if any, on monitoring if an edit was just say for a typo or it's a completely different text after the edit (and do they implement a system to auto-detect the % change and then auto-flag for manual review), do you notify users who previously viewed it or perhaps only interacted with it in some way that it was edited "x%", etc.

I assume they'll have to have things like version control, so all prior versions are readable for crowdsourced moderation, if not then that's a bad decision IMHO; some past decisions of theirs haven't been rational, and seem more fear based - so maybe they won't even have version control in the name of "privacy" - pretending that the person didn't previously said what they said, and that somehow it's deleted off the internet and everyone's memory if they delete the tweet.

Could lead to new artforms too, like the account whose single tweet gets endlessly edited with updates.

Technically, it should be no harder than the "delete" button, which has been the de facto "edit" button up until now. The article itself talks about how defining the feature from a product perspective was really the hard part, and I'd say it makes sense to take that at face value. You really do want to be able to see when a text has been edited, there do need to be some limits, etc. and defining those things is not trivial by any means.

Difficulty was not the reason why it took so long. The excuse has been that they want tweets to be immutable, so you don’t share a tweet for it to be changed later.

A soft edit would be good for this. You can edit 3 times, but revisions are kept.

Still some questions persist like:

* Do you reset the likes / retweets after an edit? If not, what happens if folks start selling the ‘edit’ on viral tweets?

* What happens if the original tweet violates the TOS but then the edited version does not? Does the original tweet in violation still get persisted? Does the edit get persisted?

> what happens if folks start selling the ‘edit’

You’re onto something. I’m pretty sure that’s going to happen if edits are not limited. We already see “wow this went viral, here’s a link to my album”

I always thought it was because Twitter was still somewhat beholden to sms.

Reply on news.ycombinator.com