#dev 2024-10-15

2024-10-15 UTC
glacier, stormyyy, troojg and [snarfed] joined the channel
#
[snarfed]
[tantek] saw your post on the BF bug, thanks for that. I'm going to start sending those DMs soon, just fyi
#
[tantek]
Oooh interesting
#
[tantek]
definitely LMK what you ended up with for the copy
#
[tantek]
[snarfed] this was my last suggested iteration: https://chat.indieweb.org/dev/2024-10-12#t1728692750720700
#
Loqi
[preview] [[tantek]] πŸŒ‰ Admin notice: Due to a Bridgy Fed bug, @tantek.com's "ActivityPub actor" was deleted, notifying every following server which then unfollowed @tantek.com. It was an accident! Feel free to refollow @tantek.com at your discretion. Sorry for the inc...
#
[tantek]
…inconvenience. [This one-time DM sent by Bridgy Fed on behalf of @tantek.com. Bridgy Fed accounts do not yet support receiving DMs]
#
[snarfed]
yup, got it
#
[tantek]
feel free to AB test if you have a few variants in mind πŸ˜‚
#
[snarfed]
I'm happy to use your words as verbatim you want. if I was to revise, here's a draft:
#
[snarfed]
Admin notice: this account was recently accidentally deleted due to a bug in its fediverse provider, Bridgy Fed, which resulted in fediverse servers removing all of its followers, including you. We've now restored the account, so feel free to re-follow at your discretion. Sorry for the inconvenience!
#
[snarfed]
(and then your footer as is)
[morganm], bterry and GuestZero joined the channel
#
donthatedontkill
sebsel: Thank you πŸ™‚ Didn't know you were here. Down or not, thanks for the page! It's super helpful πŸ™‚
#
[tantek]
[snarfed] your edit looks good! Ship it with the footer!
#
[tantek]
has anyone tried out Giscus? A comments embed powered by GitHub discussions https://github.com/giscus/giscus
#
Loqi
[preview] [giscus] giscus: A comment system powered by GitHub Discussions. :octocat: :speech_balloon: :gem:
lockywolf, [jeremycherfas]1, [snarfed]1, [morganm]1, [KevinMarks]1, [Joe_Crawford]1, IWSlackGateway2 and greybeard joined the channel
#
catgirlin.space
[snarfed] my most recent posts on my website aren’t showing up in bsky, but they do on fedi. any ideas?
#
Loqi
[preview] rosalina saige
#
[snarfed]
some posts that BF emits over its firehose to the Bluesky relay aren't getting ingested and processed. not sure why yet
GuestZero and GuestZero_ joined the channel
#
[tantek]
curious what people here (as devs) think of the "Maker Taker" description/system that Dries came up with: https://dri.es/balancing-makers-and-takers-to-scale-and-sustain-open-source
GuestZero and [Murray] joined the channel
#
[Murray]
I agree with the analogy. Feels almost self-evident and well labelled. Not sure I agree with a lot of the economics discussed, but that's another matter πŸ˜…
bret joined the channel
#
[tantek]
Yes, agreed, the aspects are separable. Appreciated [Murray]!
GuestZero joined the channel
#
capjamesg[d]
I have a question. My document store is thread unsafe, in that if I run it in multiple threads I'll end up needing to index all the data in each thread. But I want to use it with multhreaded Flask. What is my best option?
#
_pi_r2_0[d]
capjamesg[d] how is it thread-unsafe? you can have a multi-reader, single writer lock, or you can do all the lookups and mutations in a dedicated thread, or you can make it thread-safe by storing in an MVCC datastore
#
capjamesg[d]
It's thread-safe in that I haven't written a lot of multi-threaded code.
#
capjamesg[d]
Assume you can't write to the index, after loading it with data all it's doing is get() operations on dictionaries.
#
_pi_r2_0[d]
pure reads are always thread safe in the absence of mutations
#
capjamesg[d]
Oh, wonderful.
#
_pi_r2_0[d]
to get to thread safety in the presence of mutation, the simplest thing to do is to bring something like https://pypi.org/project/rwmutex/ to the equation, to take a read lock on reads and a write lock on writes
#
_pi_r2_0[d]
[edit] to get to thread safety in the presence of mutation, the simplest thing to do is to bring something like https://pypi.org/project/rwmutex/ to the equation, to take a read lock on reads and a write lock on writes
#
_pi_r2_0[d]
only one writer is allowed, if there is no writer infinite readers are possible. so make your write sections as small as possible by doing as much work as possible before (rarely, after) the shared state mutation
#
capjamesg[d]
I am using a tool called waitress to host a uwsgi server.
#
capjamesg[d]
I think it may only be running threads for requests, so I'm okay.
#
capjamesg[d]
The big thing this is in aid of is how do I reduce latency on my search engine πŸ˜„
#
capjamesg[d]
Queries are < 1ms, but it takes 50-80ms for data to get to my computer πŸ˜›
#
capjamesg[d]
And if there are lots of queries coming in, things get a bit backed up.
gRegor and tanoolh joined the channel
#
capjamesg[d]
(Thank you for all your help by the way!)
#
_pi_r2_0[d]
you might want to debounce the input events
#
_pi_r2_0[d]
instead of running a search for every keypress, send a search when there hasn't been an event for ~200ms
#
_pi_r2_0[d]
as to 50-80ms latencies… yup! what's the ping between your machine and your server?
#
capjamesg[d]
I think I may have disabled the ping ports πŸ˜…
#
[snarfed]
[tantek] DMs are going out now
#
[tantek]
whee! we'll see if any of the solicit any @-rage
#
[tantek]
any of them*
#
[tantek]
[snarfed]++ amazing of you to automate such a one-off (hopefully lol) thing
#
Loqi
[snarfed] has 53 karma in this channel over the last year (100 in all channels)
GuestZero and lazcorp joined the channel
#
[tantek]
ok so I'm tempted to flip my @-mention autolinker to use xcancel(.com) rather than twitter(.com). Any reasons I shouldn't do that? (like literally for everytime I @-mention someone in all posts 2010 to present)
[aciccarello] joined the channel
#
[mattl]
maybe change it to use http://tantek.com/twitter?foo=bar and then you can just change it in one place later?
#
[tantek]
it is one place right now, in the CASSIS auto_link function
#
[tantek]
[snarfed]++ thanks again, seeing tons of refollows πŸ₯Ή and also for making it all possible BridgyFed++
#
Loqi
BridgyFed has 8 karma in this channel over the last year (11 in all channels)
#
Loqi
[snarfed] has 54 karma in this channel over the last year (101 in all channels)
#
[snarfed]
[tantek] you're welcome! really nasty bug, apologies again, thanks for being so understanding
#
[snarfed]
on the plus side, you're the only active user it hit. on the minus side...you're the user it hit πŸ˜†
to2ds joined the channel
#
[tantek]
[snarfed] certainly made for a good real live test case! πŸ˜†
#
[tantek]
and now we have some brainstorming on how (and should) infrastructure send admin-related DMs
#
[tantek]
that's probably worth documenting as an example tbh
#
[mattl]
snarfed: if you can tell all of Tantek's followers to follow me too that'd be great πŸ˜‰
#
to2ds
[snarfed]++ - For debugging AP stuff. Took me over half the day figuring out why my homegrown AP server wasn't doing attachments correctly 😩
#
Loqi
[snarfed] has 55 karma in this channel over the last year (102 in all channels)
#
[snarfed]
to2ds I feel your pain!
[jamietanna] joined the channel
#
to2ds
[snarfed] - πŸ€œπŸ€›
#
to2ds
All of Tantek's followers, follow [mattl]
#
to2ds
Hope it works :)
#
[tantek]
maybe I should do a Follow Friday post lol
#
[tantek]
[snarfed] minor issue, does the "followers" count (on the BF profile dashboard page) need to be reset as of 2024-09-22 and/or recounted somehow?
#
[tantek]
I'm assuming I still have a subset of the prior number (<1500) whereas it looks the like the number is only going up
#
[snarfed]
[tantek] yeah I haven't cleaned those up yet
#
[snarfed]
I'm a bit reluctant because I can't be sure how different implementations handled the actor delete. I know Mastodon severed following relationships, but others may not have, so ideally I'd keep delivering to them just in case
#
[snarfed]
and I don't currently sniff server software, so I'm not easily set up to distinguish
#
[tantek]
I’m guessing 90%+ are mastodon so I'd be ok with a onetime follower count reset as of 2024-09-22, even if that undercounts other implementations that were not impacted by the bug
#
[snarfed]
I was thinking more about keeping delivery to old non-Mastodon followers where it might be working
#
[snarfed]
not so much the count
#
[snarfed]
do you have a preference? keep the old stale followers possibly working, and the stale count? or reset and possibly cut off those old still-working followers?
#
[tantek]
is that a followers vs instances question?
#
[tantek]
oh wait, [snarfed] are you saying that if someone refollows me that they don't add to the overall count because the BF database already thought they were following me?
#
[tantek]
like is BF double-counting follows from before bug and now or are those de-duped automatically?
#
[snarfed]
no, it's not about count, it's about non-Mastodon instances that may have preserved your followers, so when you post now, those old pre-delete followers might be getting your posts again
#
[snarfed]
I can reset all of your pre-delete followers - which will also "fix" your follower count - but that would cut off any of those non-Mastodon followers who are now working again and haven't re-followed
#
[snarfed]
(but yes, separate from that, BF de-dupes followers, so if someone re-follows you, they won't add to your current count since it already includes them)
#
[tantek]
Ok, given that refollows are not inflating the follower count, let's leave it as-is and see if it sorts itself out to be vaguely (within 10%) accurate
#
[snarfed]
πŸ‘
#
[0x3b0b]
Some of Tantek's stuff showed up for me again from Sunday, to provide that data point. My updated follow request still shows as "follow request sent" rather than "already following" but I don't remember what it looked like before.
#
[tantek]
0x3b0b odd, because all follow requests should be (nearly?) instantly approved AFAIK from BridgyFed
#
[tantek]
feel free to unfollow / refollow and see if that fixes it?
#
[0x3b0b]
Hmm, worth trying. Could also be a side effect of my direct actor record editing. Unfollowed; waiting a minute to make sure the new follow request gets a new ID and doesn't throw an error instead... (that's a me thing)
#
[0x3b0b]
maybe someday I will decide that only being able to act once per minute is enough of a hassle that I want to make my IDs one character longer
#
[0x3b0b]
Looks like the same result. 🀷
[0x3b0b], [schmarty], [benatwork] and superkuh joined the channel