#dev 2023-10-07

2023-10-07 UTC
gRegorLove_, sp1ff and geoffo joined the channel
#
[catgirlinspace]
new idea. i just, do relmeauth on my own site for auth. except itd just be github oauth. no passwords to store then and i dont have to learn how to use webauthn which makes like, 0 sense...
#
[tantek]
looks good!
#
aaronpk
And this is why RelMeAuth is a thing!
gnoo_ joined the channel
#
[catgirlinspace]
yep just gonna go with that for now :3 is a lot easier
geoffo, gxt_, gRegorLove_, jonnybarnes, Renfield, mretka, gerben, gRegorLove__ and bterry joined the channel
#
[snarfed]
morning all. so, afaik follow posts (https://indieweb.org/follow#How_to_publish ) are rare, most of us follow inside readers without posting to our sites, but they do happen
#
[snarfed]
unfollow posts are even more rare. we've discussed 410ing the original follow post, or publishing a separate u-unfollow-of, https://indieweb.org/unfollow#Brainstorming
#
capjamesg
appears emerging from a wild Pokéball.
#
capjamesg
angelo is interested in follow posts.
#
[snarfed]
any thoughts? BF currently supports 410, and it has a request for u-unfollow-of also, but I'm inclined to say no, that plus a UI for it is enough
#
epoch
[snarfed]++
#
Loqi
[snarfed] has 105 karma in this channel over the last year (159 in all channels)
#
epoch
for putting the space after the link and before the )
#
capjamesg
I think a lot of the discourse is around blogrolls / lists of blogs vs. individual follow actions on the wider web.
#
[snarfed]
hah yes!
#
capjamesg
I think a discrete follow action makes sense. Are people publishing "following / subscribed to" pages on their site?
#
capjamesg
Meta point: why did the web settle on the "follow" lexicon?
#
[snarfed]
probably came from feed readers and then popularized by tumblr, twitter, etc
#
capjamesg
I like how you can follow a feed without necessarily telling a site you are explicitly following their content.
#
capjamesg
(i.e. if 10k people use feed reader A and they use the same user-agent for retrieving content, the content author cannot identify each one)
#
aaronpk
The more I think about it the more I think follow posts are the wrong model
#
[snarfed]
or, maybe following in a reader is just a different thing than publishing an explicit announcement that you're following someone
#
[snarfed]
both useful, maybe, independently
#
aaronpk
public content should be public content and you can read it without the publisher knowing who you are. If you want to publish content for a specific audience, you can give them a way to authenticate, at which point you should know that they are fetching your content
#
aaronpk
blogrolls are a good way to share links to things you're following, but that is different than a follow post and doesn't need to overlap
#
[snarfed]
agreed, following in a reader is internal to the reader, no need to publish anything. I guess then I wonder how much blogrolls and follow posts should/could overlap? could a blogroll just be an h-feed of h-follow-ofs?
#
aaronpk
Adding structure to a Blogroll seems useful, and that sounds like a reasonable way to publish it in HTML instead of OPML
#
[snarfed]
this also implies that tools like Bridgy Fed shouldn't actually expect to handle follow/unfollow via webmention + mf2. following in readers or BF would generally only be in their UIs
#
[snarfed]
interestingly social networks kind of take the opposite tack. following someone both follows them (ie adds them to your timeline) _and_ shows them in your following list (equivalent to blogroll/list of follow posts), which is public if your account is public
#
[snarfed]
ugh this kind of functionality analysis and synthesis is not my strength, needs [tantek] [KevinMarks] and many of the rest of you all
#
[KevinMarks]
Follow was later - subscribe was the original verb for feeds. Follow was by contrast to Friend which needed approval by the person.
#
[KevinMarks]
We have struggled with it historically - I think the original xfn discussion suggested diffing the follows over time to see changes.
#
[KevinMarks]
It was added as a verb in AS1 because there was a common pattern of sites publishing follows in the users feed
#
[KevinMarks]
AP took that follow verb and made it imperative rather than reporting.
#
[snarfed]
also AP needed it architecturally, for the protocol, independent of any user-facing terminology or features
#
capjamesg
+1 re: AP.
#
capjamesg
I am unsure whether mf2/webmention following is in scope for BF right now given the lack of people publishing / interested in processing follows outside of AP.
IWSlackGateway, [snarfed], [snarfed]1, [aaronpk], [aaronpk]1 and [aaronpk]2 joined the channel
#
[aaronpk]
sorry [snarfed] your last 2 messages didn't go through the gateway because I was messing with it
IWSlackGateway joined the channel
#
[0x3b0b]
I think there are a few aspects of BF where the decisions are likely to be different depending on whether you are trying to provide Least Surprise for the side being bridged to or the side being bridged from...
[schmarty] and [tantek] joined the channel
#
[tantek]
[snarfed] is right that both models are likely useful, for different use-cases. And there are more like the Dopplr model
#
[tantek]
I think it would also help to document the various friend follow request-follow allow/follow UIs across silos and software
[catgirlinspace], [aaronpk], IWSlackGateway and [snarfed] joined the channel
#
[snarfed]
0x3b0b yup! bridges are an exercise in compromise
#
[snarfed]
[for the bridge] i'm idly considering a cross-protocol https://indieweb.org/SWAT0
[catgirlinspace], IWSlackGateway and [aaronpk] joined the channel
#
c​apjamesg
Oooh!
geoffo, IWSlackGateway, [aaronpk], [aaronpk]1, obi-one_kenobi, moodydot, [catgirlinspace] and [snarfed] joined the channel
#
[snarfed]
ok so I bit the bullet and briefly tried to learn JSON-LD. I took away these three ideas:
#
[snarfed]
• external schemas, referred to by URLs, via @context
#
[snarfed]
• compaction and expansion
#
[snarfed]
are those the core points? did I miss anything?
#
[snarfed]
• schema mapping between local/proprietary schemas and public ones, via @vocab
#
Loqi
misses anything too
#
[catgirlinspace]
what does json ld even _do_? i just see funny urls in it that do nothing?
#
aaronpk
i was not aware of @vocab
[tantek], geoffo and dustinm` joined the channel; obi-one_kenobi left the channel
#
Soni
we did add a link to web+ap://tag:FediLinks@domain.invalid to our website
#
Soni
seemed reasonable now that we're trying to push for tag links
#
Soni
sadly nothing supports it yet
#
[tantek]
aaronpk @vocab is how LD does mapping between different property names (URLs) that have the same semantics I believe so that a "full LD processor" can treat them as the same thing in memory or something
#
aaronpk
i thought that was @context
#
[tantek]
No @context maps short property names into long URLs
#
aaronpk
> @vocab is used specifically to declare a default vocabulary from which all terms derive, without having to declare specific mappings for each term (as you normally do in a @context).
#
[tantek]
Yes vocab can be used for that too
#
aaronpk
i really want a way to just upload a photo from my phone and have it post on my website with the date from the metadata on the photo
#
[snarfed]
extracting and using date from EXIF sounds doable entirely inside your own web server, doesn't really need any standards anywhere other than EXIF, right?
#
aaronpk
i might do it as a one-file micropub client I can post to from a shortcut
#
aaronpk
I know Shortcuts can also extract the date but it always takes me longer to create and debug shortcuts
#
Zegnat
I think petermolnar is/was displaying a dump of the photo’s EXIF data next to it on his site? Looked pretty straight forward
tabi, tabi1 and aaronpk_web_test joined the channel
#
[catgirlinspace]
i wannya do something where i can upload pictures to my website and it stores the original + a version where gps is removed, then display the safe version publicly along with a little display for the other exif data for like, camera used and stuff.
tabi2, tabi3 and tabi4 joined the channel
#
GWG
I would do what you said as well, but how would I tell the Micropub endpoint to behave that way? Custom instruction?
#
aaronpk
why would you not want it to always do that? also it seems not a feature of the micropub endpoint but of the website when it displays the photo
#
aaronpk
like, store the original, and display a subset of the info. similar to how you often want to display a scaled down size
IWSlackGateway, [aaronpk] and aaronpk_web_test joined the channel
#
GWG
aaronpk: If I upload a photo it stores that information in my site already... but it doesn't set the publish or location property of a post based on it
#
GWG
I was talking about an instruction to set post properties from the photo property
#
aaronpk
Why do you need an instruction? Why not do it always?
#
GWG
I suppose...if it is a photo post.... I could
#
GWG
I suppose if there is a conflicting property I could consider the Micropub property to override
#
GWG
I need to map out the logo
#
GWG
logic I guess
geoffo and [catgirlinspace] joined the channel
#
[catgirlinspace]
GWG personally i was planning to kinda just, do it without micropub and have my own ui on my website ig...
#
aaronpk
oh right, exif dates don't have timezones
[tantek] joined the channel
#
aaronpk
oh nice, apple puts the caption of the photo into the "ImageDescription" exif tag
#
aaronpk
ah and they added a custom tag for the timezone offset
#
aaronpk
UndefinedTag:0x9010
#
aaronpk
ooh SubjectLocation
#
aaronpk
looks like it represents a rectangle of the main subject of the photo
[schmarty] joined the channel
#
[schmarty]
You're getting close to my ideal checkin client! Use time and location data from one or more photos, reverse-geocode for address, optional comment and tags, tadaaaa
#
aaronpk
the reverse geocode is the messy part
#
[tantek]
only for new venues presumably. if you've checked in there before, you can likely re-identify when you're at the same venue and skip all the reverse geocode hoohaw
#
aaronpk
well that just means you're reverse geocoding against your own location database
#
aaronpk
and that still has problems in dense areas, but does help a lot
#
aaronpk
for a while i was doing auto-checkins, if I was in a location for more than 10 minutes it'd check if there was an unambiguous place I had been to previously and check me in automatically
#
[tantek]
heck, if you're following friends checkins, cache all their venues as well, and use those as an L2 location cache even for first time checkins at venues
#
[tantek]
my guess is that would drop the number of new to you and all your friends venues checkins down to <1% of all your checkins
#
[tantek]
beyond that you could add an L3 location cache by preloading venues adjacent to venues you've been to before, or for any street segment with 2+ venues you've checked into, preload that entire segment of venues into L3
#
[tantek]
having your prior venues (L1), your friends prior venues (L2), and pre-loaded adjacent / same-street venues (L3) in a cache means you can load that cache clientside and eliminate network latency for my guess is 99.9% of checkins, especially in urban areas.
#
[tantek]
this is useful because many "free" reverse geocoder services have API call frequency limits
#
aaronpk
based on my auto-checkin experience, I completely agree that using a local venue cache first will give you a huge benefit for the use case