Not sure how to give a concrete repro for this, but I often get a red dotted-line "new" indicator for threads which do not have anything new in them since the last time I checked. This happens often enough that I'm pretty sure it's not because people are making minor edits or something that I just don't notice...

- Casey

I have also noticed this.
Just now I clicked on https://handmade.network/forums/t/1338/2#8355 thread. Went back, pressed F5, still shows NEW! Either it is some caching, or a bug.

cmuratori
Not sure how to give a concrete repro for this, but I often get a red dotted-line "new" indicator for threads which do not have anything new in them since the last time I checked. This happens often enough that I'm pretty sure it's not because people are making minor edits or something that I just don't notice...

- Casey

mmozeiko
I have also noticed this.
Just now I clicked on https://handmade.network/forums/t/1338/2#8355 thread. Went back, pressed F5, still shows NEW! Either it is some caching, or a bug.

Thanks guys. I've just had it happen myself. If I leave the rest of the threads alone, I should hopefully have a concrete case to repo it on my development machine and see what's going on here.

It's a surprisingly difficult thing to track last read information for 8000+ posts in ~1500 threads for 1500+ people and not keep track of and have to update essentially way too much data, or else have to put synchronisation in all over the place which kills performance.

The algo I came up with is at its heart very simple, stores things sparsely and does away with many of these concerns. It could however be that we're seeing a poor interaction with some caching issue or that I've overlooked an edge case somewhere. This bug didn't show up during my local testing before I pushed it live. In any event, I'm going to look at the data for my account keeping track of this information and see why the Odin thread's unread status seems to be off, to start with.

Edit:
I'll write a blog post this month with why this is a surprisingly tricky feature to implement efficiently, how I implemented it, and why some of the even more efficient ways to implement it have drawbacks of their own that outweigh their benefits.

Edit 2:
I've found a repro case, so I think I can probably isolate and fix this bug in fairly short order.

This seems to have been fixed. Thanks for the report, guys.