interface.png

@amerika @Cookieman @Looming @NEETzsche @childrapist1488 @fedilist @lina @mint It is true that the interface shapes things, but it is not the case that the same interface means the same use-case, the same intent, the same audience, the same experience, etc. The interface started off as mostly Twitter-like (though both have diverged), but this place isn't anything like Twitter. And I can swap out the interface and mostly have the same experience, so the interface is fungible.

If I use regular PleromaFE, or if I use bloat, or if I use the local hacked BBS interface (diverged from the one that used to ship with Pleroma, see https://git.pleroma.social/Duponin/sshocial for a server-independent version, here's a screenshot with some posts and the help message), the same things happen, more or less: I'm talking to the same people about the same topics regardless. There are very few differences: some extra interaction required to see an image (copy-paste in Linux or, just plumb the URL in Plan 9, in which I have installed the very fetching and surprisingly readable Fixedsys font from Windows 3.1), things like that.

It is conceivable (and I may set myself to the task later, as I have been threatening to do for a few years) that you get the server to speak IMAP to allow someone to view messages and then SMTP or submission to send messages (the metadata is nearly 1:1, down to the In-Reply-To header). In that case, the interface would be identical to any other email client, because you would be using your email client. I do not think you could correctly call the system "email", though, any more than you could call this "social media" because of the superficial resemblance to Twitter. It certainly would shape some of the interaction, but I can't imagine that I'd use it the same way I use regular email.

> The biggest difference here is no algo, which has both positives and negatives.

There's really nothing that stops someone from implementing a non-chronological timeline (hopefully in parallel with the existing timeline rather than as a replacement), except that the predatory version of it has given everyone an allergy to the concept. Instead of sorting by date, you calculate a ranking, you use the same formula finance guys use for discounting future dollars, you sort by the ranking. (`value/((1+depreciation_constant)^age)`, use positive interaction count for value, tweak the depreciation constant until you are happy with the rate of drop-off, or if you wanted to get fancy, you could depreciate the interactions themselves, so an old post that gets a lot of very recent attention has a chance of bubbling up. For news articles or something, depreciation by day is generally better but on fedi you could do it by the hour.) You could even do it pretty quickly: a sparse table with the rank, join it for sorting the special timeline. Posts that go under a given threshold get removed from the sparse table, and the rest of them you could just periodically depreciate (which would make the "fancy" version easier to implement than the other version). So if you cap the sparse table at maybe a thousand entries and you could actually afford to process it in real-time. I haven't looked, but I suspect this is more or less what HN does; it's what anyone does if they have a math nerd (or a sufficiently clever finance nerd) working there.

Anyway, not hard to do, just no one has done it. (Mastodon has something similar, but they use absolute numbers. Twitter figured out that you have to account for ambient level for that to be useful, after some amount of time with "Trending" tab containing exclusively Bieber-related hashtags. If you are trying to get the computer to tell you how today is different from yesterday, you have to ask to about the proportional change from yesterday rather than just looking at the top.)
interface.png

Public

Notices where this attachment appears