#social 2019-02-01

2019-02-01 UTC
#
dansup
cwebber++
#
Loqi
cwebber has 3 karma over the last year
#
cwebber
encrypted p2p-deliverable demo of activitypub stuff works :)
#
jdormit
cwebber: Wooo, congrats! Great news for you and AP
#
jdormit
What did getting that grant involve?
#
cwebber
thanks jdormit :)
#
cwebber
jdormit: filling out an application (not hard) and several years of mostly unpaid labor on standards that got me some noteriety (hard) :)
#
jdormit
Haha fair enough
#
jdormit
Glad to hear you're finally getting some funding :)
xmpp-social, _xmpp-social, cwebber2 and Guest84 joined the channel
#
fr33domlover
cwebber2, how does it store stuff? ipfs? torrent? something else? and who fetches the content, the AP client (by downloading torrents etc.) or server (by offering the content to its users via regular http)?
#
JasonRobinson[m]
cwebber2 (IRC): wow pretty epic, congrats!
#
cwebber2
thanks JasonRobinson[m] :)
#
cwebber2
fr33domlover: the demo does the simplest thing possible to demonstrate things clearly just to be a simple demo. there's no AP C2S so I could focus on S2S only, the web server also acts as a minimal client. The way it saves it is by saving the content-addressed encrypted "chunks" to a local store, but then if it doesn't have the chunks, it has a preconfigured set of peers that it looks for the chunks there and saves them locally
#
cwebber2
it's only 500 lines, I wanted to try to boil the concept down to its most basic
timbl joined the channel
#
jdormit
so here's my current AP conundrum
#
jdormit
if I receive a follow and want to auto-accept it, am I creating a race condition on the server that sent the follow if I send the Accept within the same request where I received the follow?
#
jdormit
e.g. suppose the sending server only handle Accepts for which it knows it sent the follow
#
nightpool[m]
the idea of doing any remote request inside of a different request creeps me out
#
jdormit
although I would guess the sending server would persist the activity before attempting to deliver it, so maybe I am overthinking it...
#
jdormit
yeah PHP is garbage
#
nightpool[m]
how do you know the remote server won't just timeout?
#
jdormit
I'd send the request async
#
jdormit
but I'm still worried about the possibility that the sending server receives my Accept before it receives my response to the Follow
#
nightpool[m]
I'm less concerned about that.
#
jdormit
yeah I may be overthinking this
#
nightpool[m]
ideally they've committed the follow to the database before sending it, and if they haven't then that will be a fun bug for you two to figure out but I don't think it's very likely
#
jdormit
haha yeah
#
jdormit
I really don't like PHP's cgi-style request execution
#
jdormit
the script only lives for the duration of the request, so everything needs to get done within that context
#
jdormit
or you do something janky like send a second request to your own server from within the script to execute some async action in another request
#
nightpool[m]
?
#
nightpool[m]
most php apps I know use a job queueing system
#
nightpool[m]
at least, the federated web ones
#
jdormit
yeah, that would be ideal but I am writing a library and targeting WordPress
#
jdormit
so it can't rely on any queueing services
#
nightpool[m]
😬 ah
#
jdormit
hmm reactphp looks promising maybe
#
jdormit
it says it doesn't require any extensions
#
jdormit
well, seems like I'll be safe-ish doing async requests. I'll do it that way and see if anything breaks down the line 🤷
timbl, jdormit, cwebber2 and cwebber joined the channel; jdormit left the channel
#
cwebber
hello from the road
dustyweb_, cwebber2, DenSchub, Luke_ and ma1uta joined the channel