if we don't have reactions with their own kind, then people will use kind:1 to write a single emoji without a way to single them out. are kind:1 notes with one or 2 characters spam or wasteful?
at least this way we can choose to not download reactions if we don't want to (or zaps or any other kind).
now, regarding note aggregation, this is a practice that doesn't work well with nostr, because there is no central database. every relay has it's own count and trying to merge all this is pointless, what value do people get from an open count? this number can be faked or inflated and is only used by people manipulating the masses by bragging their own numbers. notes with 3 / 4 likes sometimes come from the same account that made the post in the first place, and because clients aggregate and don't show who did what and when, they distort the value of the note and burden the relays. that's why advocating for more of this is going the wrong way, clients will just show meaningless numbers to people.
> this number can be faked or inflated and is only used by people manipulating the masses by bragging their own numbers. notes with 3 / 4 likes sometimes come from the same account that made the post in the first place, and because clients aggregate and don't show who did what and when, they distort the value of the note and burden the relays. that's why advocating for more of this is going the wrong way, clients will just show meaningless numbers to people.
good points but I would argue you can always create new npub bots to like your content and then it would not matter if aggregated or not, the numbers would be inflated. Sure relays will have not the full sight but it would be a fast way to receive an initial count without wasting too much time downloading all reaction notes. Clients could fetch that and after fetching more important stuff get the notes so it would just speed up things. This all assumes that reactions are not that important and it's ok to delay downloading of the full data to speed up ux.