#dev 2021-04-17

2021-04-17 UTC
#
jacky haha
Seirdy and [tantek] joined the channel
#
jacky TIL that the whole field[]=3&field[]=4 is a nonstandard thing
#
jacky for like the URL standard, that is
#
aaronpk Yup
#
aaronpk technically it's allowed to have repeating URL params but many frameworks don't allow that
#
jacky learned that the hard way today lol
#
jacky via https://github.com/seanmonstar/warp/issues/733
#
jacky er https://github.com/seanmonstar/warp/issues/733#issuecomment-720166509
#
Loqi [ZettaScript] #733 Not compatible with HTML checkbox lists
#
Loqi [Gelox] As far as I can tell `serde_qs` does not have the same list syntax for query strings that is usually the case for web frameworks, at least from the two sources I posted in the other issue [stackoverflow](https://stackoverflow.com/questions/3061273/se...
#
jacky hyper annoying
#
[tantek] GWG++ ENHANCE++
#
Loqi ENHANCE has 1 karma over the last year
#
Loqi GWG has 18 karma in this channel over the last year (130 in all channels)
#
jacky re: repeating URL params and my 'shock' https://imgur.com/a/znPiXbK
#
GWG [tantek]: Enhance gets a karma point?
#
[tantek] GWG, it was a bit of a ++ joke 🙂
#
[tantek] on the whole ENHANCE trope
#
GWG [tantek]: To be honest, webmentions is slow going.
#
GWG The code has so much baggage
#
[tantek] GWG, is that the primary blocker for merging Webmentions & Semantic Linkbacks plugins so folks don't have to be confused like that one person that recently tweeted?
#
GWG [tantek]: Time is one too.
#
GWG [tantek]: We get stalled a lot on it
#
[tantek] really Webmentions plugin should absorb 100% of the Semantic Linkbacks functionality, and then SL should be retired
#
GWG We have a whole bunch of new parsing code merged into a branch of webmentions.
#
[tantek] that way "Webmentions" plugin can actually give people what is popularly known as "Webmentions", i.e. cross-site comments, likes, etc.
#
GWG After parsing code, the post type discovery code has to be built up.
#
GWG Then, we have to migrate the old data from semantic linkbacks
#
[tantek] can you reframe that to response type discovery code instead?
#
[tantek] you don't need all of post type discovery
#
[tantek] there is a separate smaller algorithm for response type discovery, deliberately for this reason
#
GWG [tantek]: You are correct. I was thinking response type.
#
GWG I misspoke.
#
[tantek] 👍
#
GWG But the biggest problem is the data from Semantic Linkbacks to migrate over
#
GWG Not even at that point though
#
GWG Either way, the biggest issue is too many projects, never enough time.
#
[tantek] that I'm far too familiar with 😕
#
GWG I try to pollinate many projects at once
GWG, [jeremycherfas], minoru_shiraeesh, jjuran_, silo_, jdmdm and jdmdmjsnd joined the channel
#
jeremycherfas That would be because you are a busy bee, GWG
dhanesh, [suze_shardlow1], minoru_shiraeesh, sebbu3 and [jeremycherfas] joined the channel
#
[jeremycherfas] !tell GWG is this of possible interest https://github.com/ideditor/country-coder
#
Loqi Ok, I'll tell them that when I see them next
#
Loqi [ideditor] country-coder: 📍➡️ 🇩🇰 Convert longitude-latitude pairs to ISO 3166-1 codes quickly and locally
#
GWG Interesting.
[tantek], [chrisaldrich] and [manton] joined the channel
#
[manton] For folks implementing ActivityPub, any tips on when to give up on a Mastodon instance that has gone away? Thinking about removing any “follows” from an instance that can’t be reached after some period of time, or number of failed connections.
KartikPrabhu joined the channel
#
aaronpk i should really do that too
#
Loqi definitely
#
GWG [manton]: I remember snarfed talking about this for webmention endpoints a while back. Something about a gradual backoff process.
#
GWG That you gradually reduce the checks until you declare it gone.
#
jacky yeah ^
#
jacky exponential back off attempts
#
jacky (or Fibonacci)
#
jacky on both the connection timeout _and_ response timeout
#
[manton] That makes sense, thanks y’all. I think I might start with just counting the continuous failed attempts so I can have some more data to figure out when an instance is probably gone and not coming back.
[Rose], shoesNsocks, [KevinMarks] and [snarfed] joined the channel
#
[snarfed] oh yeah GWG bridgy’s backoff is only per webmention, not permanent for an entire host
#
GWG [snarfed]: I was speaking philosophically, not identical implementation
#
[snarfed] ok!
#
[snarfed] yeah, i don’t have anything automatic for this specifically, i just occasionally notice dead mastodon hosts and manually disable their users. bridgy’s semi-secret admin interface helps: https://brid.gy/admin/sources , https://brid.gy/admin/responses
__minoru__shirae, [tantek] and shoesNsocks1 joined the channel
#
@jaybearca ↩️ Thanks a lot, Chris. http://Brid.gy sounds like a great solution to this. I’ll check it out. ☺️ (twitter.com/_/status/1383515738250485762)
Seirdy and [manton] joined the channel
#
[manton] Thanks [snarfed] for Bridgy Fed source and issues on GitHub! Helped me debug some ActivityPub stuff today after hitting my head against the wall for hours.
[KevinMarks] joined the channel
#
[KevinMarks] https://seirdy.one/2021/04/16/permissions-policy-floc-misinfo.html
#
Loqi [Rohan Kumar] Misinformation about Permissions Policy and FLoC
#
Seirdy hey, it me!
#
sknebel I honestly do not understand where you get the info from to confidently state that sites that do not do the JS call are not included in the calculation
#
sknebel but looking at the source code, it appears to be correct
#
sknebel sigh, crappy documentation
#
Seirdy sknebel: i said that sites that do not include the js call do not *leverage* FLoC. I addressed cohort calculation in a dedicated section.
#
Seirdy Impacting cohort calculation and actually utilizing FLoC are two different issues
#
sknebel okay, I think I get your point, although opting out of cohort calculation is the only thing I've seen people recommend that header for
#
Seirdy for cohort calculation, opting your site out may or may not impact which cohorts get created and which cohort a user is assigned to. Changing which cohort a user is assigned to may or may not make their fingerprint less unique. So the opt-out header is kinda meh ¯\_ (ツ)_/¯. people have been trying to get upstream software to implement this band-aid solution in a massive knee-jerk
#
Seirdy reaction.
#
Seirdy and this tiny band-aid depends on Google's non-standard usage of the permissions-policy header, which they can stop doing whenever they want (see Do Not Track). This is why FLoC is terrible: it's so hard to fight against it.
#
Seirdy The only winning moves are a) not to leverage it on your sites (stop using 3rd-party adtech), and b) to get people off Chrome
#
Seirdy (and derivatives that have stuff like this enabled)
[snarfed] joined the channel
#
[snarfed] [manton] hey, you’re welcome! glad it helped