So when you share something on mastodon, even if no one clicks a link, it is going to get Ο(n) connections from n servers fetching for previews.
If you have n followers, even if many of them stopped logging in a year ago, your server is going to need to do Ο(n) connections to feed out that information.
If you want to load a tree of replies, you're looking at Ο(n) requests.
These are all incredibly expensive burdens and operations that should not be that burdensome.
One thing about larger systems and more centralized systems is that they _don't have these problems_.
Not have more ways to work around these problems. They _flat do not have them_. To the degree they do have them, they can often solve them in significantly less time and with less complexity than solving them in mastodon takes.
If you believe in a vision for the fediverse that involves a lot of small servers, this should worry you and is something to spend some time on.
I'm kinda embarrassed I know so little of echomails implementation.
It wasn't small; some 35,000 nodes ("servers") by the end, a half million users.
@mauve @hrefna
Echo Mail in FidoNet did that. FidoNet was a dumb point to point, hierarchical store and forward system, with Echo Mail carried in "bundles" that were decomposed at a node, a hash or something was looked at to see which items in the container were seen before, new ones made available locally.
I wasn't part of echo Mail so I don't know the details. But it was extremely robust, knew to avoid loops, was self healing regarding dropped bundles etc.