hhmx.de

· Föderation EN Mi 16.10.2024 20:33:56

For example, here's the notification UI for a very similar post on Bluesky (left) and Mastodon (right). These are almost the same height as each other... yet one is able to make room for the image, and one isn't.

Medien: 2

Föderation EN Mi 16.10.2024 20:34:58

Now that I look at it, the Mastodon version has made room for a line showing the name of the account that made the post - which seems redundant, since this is only showing interactions *on your own post*. Removing that information provides vertical space for a small copy of the image, like Bluesky does.

Föderation EN Mi 16.10.2024 20:39:25

As an aside, I've only used the new UI on botsin.space, but does it seem *slower* to anyone else? On that instance, at least, it can take multiple seconds to load the notifications UI - it was effectively instance in Mastodon 4.2.

Föderation EN Mi 16.10.2024 20:41:59

@misty i mean i dont notice a difference here, but its slow on botsin.space bc botsin.space is very slow

Föderation EN Mi 16.10.2024 20:42:21

@pup_hime Yeah that's fair. I guess this is more of a thing where an already-slow instance performs even worse with the new UI

Föderation EN Mi 16.10.2024 20:44:05

@misty this might be a "works on my machine" thing and ALSO the fact my account at blorbo.social (also on 4.3.x) is relatively new, but the new notifs UI works at the expected speed there than on botsin.space

but botsin also just took like 45 seconds to load the actual homepage for me (before timeline content, which is just me anyway) versus blorbo

Föderation EN Mi 16.10.2024 20:45:47

@jahtnamas Yeah that's fair. So this is probably just a thing where an already-slow server is performing worse on the new UI, but a faster server is fine on either

Föderation EN Mi 16.10.2024 20:50:02

@misty The new grouped notifications endpoint may be slower, because it needs to serialise more data than before. But the P90 on mastodon.social is 600ms, so it's not *too* slow.
We know we need to improve this, but it related to JSON serialisation performance and we havent found a good way of improving it without a lot of work.