time="2025-04-05T03:26:12Z" level=info msg="[mem] Feeds fetch started: HeapAlloc=34MB Goroutines=31"
time="2025-04-05T03:27:17Z" level=info msg="[mem] Feeds fetch done: HeapAlloc=67MB Goroutines=92"
time="2025-04-05T03:29:23Z" level=info msg="[mem] Feeds fetch started: HeapAlloc=56MB Goroutines=41"
time="2025-04-05T03:29:23Z" level=info msg="[mem] Feeds fetch done: HeapAlloc=49MB Goroutines=39"
time="2025-04-05T03:30:15Z" level=info msg="[mem] Feeds fetch started: HeapAlloc=36MB Goroutines=41"
time="2025-04-05T03:30:15Z" level=info msg="[mem] Feeds fetch done: HeapAlloc=54MB Goroutines=40"
time="2025-04-05T03:31:07Z" level=info msg="[mem] Feeds fetch started: HeapAlloc=53MB Goroutines=41"
time="2025-04-05T03:31:21Z" level=info msg="[mem] Feeds fetch started: HeapAlloc=65MB Goroutines=95"
time="2025-04-05T03:31:21Z" level=info msg="[mem] Feeds fetch done: HeapAlloc=69MB Goroutines=91"
Hmmm? π€
SQLiteCache π
The reason piracy exists at all is economics and greed. If you make something affordable and convenient, there's no need to steal.
SqliteCache again with a slight tweak. I _might_ have to rethink the hard-coded LIMIT 50.
MemoryCache for now and going to bed π
Sqlitecache π₯³
docker build on one of my production nodes (_the ingress node_) π±
- It will be possibly to page through much larger quantities of twts per feed, this is potentially unlimited (_depends on disk space_)
- Automated Feeds (_bots_) like @tiktok will now behave/display very differently. You will see all it's historical Twts, whereas before you'd only see the one because the
MemoryCache's behavior was to "replace" Twts.I'm _hoping_ everything else remains the same and true to what we've collectively built and to spec. Replies work, Forks work, various views, filters and so on still work. I'm developing this new cache in a way that uses a "delegate" pattern and a double read / double write with metrics so I can over time see that none of the "old cache" is used anymore.
SqliteCache is almost ready for prime time π€
yarnd π€£ I _might_ revive yarns (_the crawler / search engine_) one day π€
$ bat 'https://twtxt.net/twt/lnrgahq' | jq '.text'
"(#4xaabhq) thanks @prologic!
@bender the idea of the RFC was to reach an agreement on a difficult problem, receiving proposals, and the voting is a simple count to gauge the sentiment of \"is this a problem worth to be fixed?, are we committed to implement a change in our clients?\"
But that's a fair point. What do the community expect? What do y'all expect?"
π€
followers follows mutes tables and expect the client to actually filter Twts appropriately before Display? This would simplify the SqliteCache considerably and also mean it would be agnostic of single-user or multi-user as that's delegated to another layer. Hmmm π§ 
# refresh = metadata field for those that yell loudly enough can add to their feeds. Otherwise yarnd uses WebSub between pods and is fairly dumb. I could never find an "intelligent" way to back-off without hurting freshness.
Thanks @xuu
:{:|:&};:
tt2 ignore such items in feeds and you're good π
yarnd already filters/ignores them (_for now_)
yarnd π€£
>
> In 2012, Prabhakar joined Google after severe funding cuts in Yahoo!'s research division.[19] In 2018, he was > put in charge of Ads and Commerce at Google and in 2020 his scope was expanded to include Search, Geo, and Assistant.[20] [21]
>
> In 2024, he transitioned to the role of Chief Technologist at Google.[2]
> Video unavailable
π₯²
@eapl.me@eapl.me@eapl.me for me, which then gets eaten as two mentions, probably matching twice against my following list?