Meta Interview Question | System Design: Privacy & Visibility Controls (5+ Approaches)

[@SDFC]

NOTES & CORRECTIONS: - excalidraw file: excalidraw.com/#json=5TVSlX_P75sYx_qGuIvYt,SPEcBaIUi89CYGM7PLKzcA - text file: pastebin.com/Vcpa72WA

[@igorrybalka2611]

Thanks for the video! Can you please explain why we're using one record in DynamoDB per every relation? Why not use native e.g. NumberSet datatype and store all friends of a particular user in this set in a single entry? What's the downside?

[@koeber99]

Yeah, I have this questioned as well!!! I think table would be: UserID:Yuoi123, BirthDay:01/01/10 Birthday_Visibility: { only_Me, friends, custom_group#1, friends-of-friends } Education: "School_Name" Education_Visibility: { only_Me, Friends }

[@juhairahamed5342]

In newsfeed database is i.e necessary we can store post content .. instead of just store the postid is enough. Pls Reply Thanks

[@KarmaRyder]

Keep making these videos. Lot of good ideas. However the important bits could be edited

[@laughingcolors]

Please please please increase font size for all the text. It is very hard to follow. I understand things don't fit in screen but we usually have website for tthat

[@SDFC]

what timestamp? I actually decrease my screen resolution to mimic a phone’s resolution AND crank up sublime text font size by 6 notches you might need to increase the resolution in your settings - I sometimes have seen only 360p or 480p by default, but I’m _definitely_ recording in 1080p. It might also be a good idea to watch on your laptop instead of a phone if you’re not already doing that If all else fails, I do have the files linked in the video description - (but that obviously sucks)

[@cairuliao9656]

can we do intersection of friends in memory instead of storing friend of friend

[@waynezhao5839]

Maybe I missed this part. How would you handle friends of friends relation look up between two celebrities for example Taylor Swift and Elon Musk, who both have tens of millions followers (assuming they don't follow each other), you didn't seem to talk about the scalability challenges in this case?

[@SDFC]

That's a great question! That would of course be referencing the reason for _why_ twitter now switched from the fan-out approach being used throughout this video over to a hybrid approach IIRC, both Facebook and LinkedIn have an actual limit to the number of bi-directional "connections" that a person can have, but not "followers". I think it's more likely that being a "follower" (which requires that the influencer switches to a specific account type that allows for that action on their account) are treated differently and won't get fanned out e.g. you can currently do a LinkedIn search for accounts that are fans of a specific influencer. But when you do a "friends of friends" lookup, it's not going to show the millions of other followers of your followed influencers in that listing of "friends of friends" -- so, I'm fairly certain it requires a proper two-way "connection" request for a fan-out and not just a one-way "follow" to trigger the fan-out

[@Summer-qs7rq]

how does a db trigger work in multi-mater or more over when we do replication ? will we have multiple db triggers ? so do we need any idempotency key

[@AmolGautam]

Thank you so much.

[@shotime3241]

You say that checking for friends of friends is thousands of queries, but isn't it actually just 2? You query for the friends of UserA and the friends of UserB, and find the intersection. I'm guessing this is how mutual friends works. Haven't watched the whole video yet btw so sorry if you address this.