@prologic does not follow you (they may not see your replies!)

Recent Twts

Recent twts from prologic

(#budayha) I thought about that for a brief moment, but I realized that most of the content we’ve created at actually centers around very specific content related to the twtxt spec itself and the extensions we’ve created atop and around it without breaking existing clients. So I’m not sure 🤔 – I think development has to be a bit different anyway, how to build the software, the UI/UX, the API, the various tools and services we’ve built to support the larger platform, etc…


Anyway… Congrats 🎉 Proud new owner of the new pod 🥳 – What’s this now, 4 pods now in the wild that aren’t my own or managed by me? 🤔 Gonna have to start figuring out a way for the Yarns search engine to sart accounting for them? Or a way to identify them as a measure of “count”?


(#736inyq) Hmmm this is particualrly bad actually from the point of view of “wtf happened” 😂 No problems mate! 🤗 I’d just love to know how to reproduce what happened here so I can fix the backend to umm not ingest things like this? Not sure actually… There’s no source feed either which is really strange 😳

#iyijmga (#736inyq) Yeah they’re gone from my timeline too, now, but they were ingested by my pod, so these will definitely collide. I’d just be interested to see what the ingested data was, so I’m building an application/json content-negotiation for permalinks as I write this.


(#5mkqwua) Perfect 👌

* 673fcb6 2021-09-23 | Add @<movq>'s client jenny as a recommended supported client (HEAD -> master, origin/master) [James Mills]


(#4hbckpq) Oh boi! 😳

$ docker images | grep libretranslate
libretranslate/libretranslate     latest    210f515faa80   9 days ago     4.13GB


Updating language models
Found 42 models
Downloading Arabic → English (1.0) ...
Downloading Chinese → English (1.1) ...
Downloading Dutch → English (1.4) ...
Downloading English → Arabic (1.0) ...

… This is a rather expensive piece of infa. How long did it take to download all the language models? Can they be downloaded and store in a volume? 🤔 The image size itself is insane 😱

#sfdn72a (#rivvvna) Did you actually have a look at the project? 🤔 It actually supports just analyzing an access.log; But I don’t think doing this really provides enough insight of your site’s traffic IHMO. My question is more along the lines of:

If I run my own analytics for my own sites like yarn.socia would most people be okay with that?


#2my4w4a (#ojybh3a) What two pieces of context do you need in the templates to make what you’ve envisioned work?

  • InReplyTo – A short string of what current post was “in reply” to?
  • ForkedFrom – A short string of what a conversation was “forked from”?

Is this all you need? If so can you give some guidance on what they should ideally look like content-wise. For the API (REST+JSON) I would just supply the hashes and nothing more.

#ppfmj3q (#ojybh3a) For example:

If you follow the permalink for the root of the conversation that got forked 4hbckpq: (#xic44sa) Could my instance of LibreTranslate help with building a prototype?


@jlj (#ywqpb6a) This is why I think it’s important that the Yarns search engine and once implemented the one at actually re-crawl active feeds more often with some sensible algorithm. I’m already tracking moving averages for “fetch time” and “new posts”, so it’ll be interesting to see those numbers soon, but if we agree this is a good idea I think it could help solve this very problem by the yarnd backend going:

Oh oops, I don’t have that hash, lemme go see if my configured search engine does.
Yes? Fetch and cache.