#dev 2023-07-04

2023-07-04 UTC
#
gRegor Does gifv have better support or is that just an imgur thing with <video> fallback?
#
gRegor Ah, it does appear to just be a <video> according to https://en.wikipedia.org/wiki/GIF#Animation_formats
[0x3b0b], geoffo, tei_, dustinm` and [jacky] joined the channel
#
[jacky] good question [schmarty]. I _could_ see /reaction being extended to include any non-text content now tbh
#
[jacky] like expanding to allow for still image reactions too
IWDiscord joined the channel
#
[jacky] working on that change now *vikanezrimaya* (though I can't log into GitLab, cloudflare is playing games)
#
[jacky] looks like they've made a breaking change in html5ever but only bumped it as a minor version 😭
#
[jacky] oh damn rcdom got dropped
#
[jacky] I'm going to have to either rewrite the parser logic (as it depended heavily on RcDom) or move it to use scraper (but that'll be fragile because it's a freeform element scanner)
#
[jacky] how are you using html5ever to do DOM transversal, *vikanezrimaya?*
AramZS joined the channel
#
[jacky] ah I see it's via dependencies
#
[jacky] how _hard_ of a blocker is this for you, vika? I might spend tomorrow looking into this
[jeremycherfas], sivoais, win0err, Aidhan, gRegorLove_ and gRegorLove__ joined the channel
#
[aciccarello] Well that's annoying. Google is reporting that two of my image only posts are the same content even though the alt text is different
#
[aciccarello] You'd think they'd understand how to parse image content
#
[aciccarello] Guessing by their example markup, you need some visible text to differentiate https://developers.google.com/search/docs/appearance/structured-data/image-license-metadata#single-image
#
vikanezrimaya [jacky]: regarding html5ever, mostly an annoyance. RcDom was really convenient and I even used it in my prototype webmention checker to check for freeform links in content
#
vikanezrimaya I wonder why RcDom got dropped
#
vikanezrimaya oh, it's because they think of it as internal-only. jacky: you might wanna consider using a fork that exposes RcDom
rhiaro joined the channel
#
vikanezrimaya it might also be that xml-rs is capable of replacing RcDom
#
vikanezrimaya servo uses it
#
vikanezrimaya some android code or something, not too sure
[Murray] and [pfefferle] joined the channel
#
capjamesg Huffman trees are magic.
#
capjamesg [snarfed] Could you share a spreadsheet of the links in the IndieMap crawl?
#
capjamesg I tried in BigQuery but I get the error "Access Denied: Project indie-map: User does not have bigquery.jobs.create"
joshproehl joined the channel
#
capjamesg I actually don't need the data anymore, but this may still be relevant.
sivoais joined the channel
#
bkil [tantek]superkuh : (going through the chat log now that I have some time) I also automatically rewrite links to old.reddit.com via Privacy Redirect. I made a UserCSS and blocked all of their assets so that I could render it instantly from a single round trip with minimal data traffic.
#
bkil vladimyr Soni (2023-07-01T21:47:57) I think fmrl reinvented tumblelogs (and microblogs) that was in turn inspired by how people set their presence status on instant messaging chat networks in the late 90s to produce a meta-channel for their network of friends, so we went a full circle!
#
bkil [jacky]: (2023-07-02T15:01:01) On the contrary, I find the threat level between web apps and various types of native apps very similar. https://bkil.gitlab.io/secuchart/#e2ee-web-apps
#
bkil [schmarty]: (2023-07-03T22:55:21) How about substituting tiny and compact vector graphics produced by HTML & CSS shapes alone or even via ASCII-art and looping it through CSS transition/animation? A JS;DR asciinema site comes to mind.
#
[jacky] Hm, alright. I do see ammonia providing a means of working with it by vendoring it into the project
#
[jacky] I'll try a few options that keep the library small
GWG joined the channel
#
[snarfed] capjamesg sorry for the trouble! yeah the public BigQuery access has been broken for a bit, even though it's set to fully public. haven't figured out why yet
#
[snarfed] you can also get the links from the API JSON files: https://indiemap.org/docs.html#social-graph-data
#
[schmarty] bkil: feel free, but I won't! converting gifs to shapes and css animation is absurdly disproportionate effort for the content in question. Additionally, CSS from a Micropub tool typically won't "survive" when the content is displayed on one's own site, much less microformats conversion and sanitization when displayed as a reply, so it doesn't address the primary issue I mentioned.
btrem joined the channel
#
[jacky] I'm curious about the idea of sending a Webmention to a _part_ of a page
#
[jacky] like if I have a note that's long, being able to select a sentence or paragraph to do so
#
[jacky] I know that Google has its approach for partial page selection but that creates a space for super long URLs and only works if the content's text doesn't change or is in the same language as published
#
[jacky] obviously these things break if the content's changed so it might be less relevant
#
[jacky] I can see the case of doing quick hashing of sentences (or giving them a incremental ID) and allowing that to be usable for this (a fragmention, I guess)
#
[jacky] what is a fragmention
#
Loqi A fragmention is an extension to URL syntax that links and cites a phrase within a document by using a URL fragment consisting of the phrase itself, including whitespace https://indieweb.org/fragmention
#
bkil [jacky]: I think that by today, everyone supports it except Firefox who wants to stick to the scheme they had invented in the past.
#
[jacky] Which scheme is that?
#
[jacky] The partial ID approach from XML?
#
bkil No, the W3C WICG scroll to text fragments: https://en.wikipedia.org/wiki/URI_fragment#Proposals
#
bkil I also think that a URL up to 72 characters long should not cause an issue anywhere at all (consider gmail wrapping if you send it without HTML-mail), and should be considered normal up to a few hundred characters. Various limitations apply, but about 8kB of URI should still work basically everywhere.
win0err, geoffo and IWSlackGateway joined the channel
#
c​apjamesg snarfed No worries! Thank you!
#
c​apjamesg I was reading a guide on calculating PageRank with eigenvectors. It was a bit over my head but I wondered if I could do it on the IndieMap dataset.
#
c​apjamesg https://allendowney.github.io/DSIRP/pagerank.html
[schmarty] joined the channel
#
[schmarty] welp, today i learned that tenor GIFs is owned by google. https://tenor.com/gifapi/documentation#apiterms
geoffo, [timothy_chambe], tei_, jonnybarnes, eitilt and [KevinMarks] joined the channel
#
[KevinMarks] An actual fragment id makes sense as a target, and there is prior art for that when you have multiple posts per page. Fragmention was designed to extend that model (which is why it selects the containing element). Google's thing is designed to highlight discontinuous text on the page not select an element
geoffo, btrem, tei_ and [snarfed] joined the channel
#
[snarfed] thought a bit again about how indieweb tools should handle posts with multiple things in them, eg http://hugo.soucy.cc/2023/mp/ovni-a-la-rtbf.html which is a note, a like, and a bookmark
#
[snarfed] the consensus answer seems to be post type discovery, which is a deterministic way to choose one and only one thing from any post
#
Loqi [preview] [Hugo Soucy] Entre mystère et secrets, la fascinante histoire des OVNIS
#
[snarfed] Bridgy does that, https://brid.gy/about#multiple-types , so I guess Bridgy Fed should too
#
[snarfed] anyone have any other thoughts?
tei_ and geoffo joined the channel