#dev 2024-10-11

2024-10-11 UTC
[0x3b0b], angelo and bret joined the channel
#
[snarfed] ok [tantek] my current theory for your Bridgy Fed fediverse account is that accidentally deleting your actor reset your followers on each remote instance, so you need at least one new follow on a given instance before it will accept and show your new posts on your profile
#
[snarfed] https://federate.social/ is an example of an instance where you have a new follower since then. search for @tantek.com@tantek.com there, it has your recent posts. (I'm guessing none of us here went there and manually searched for them)
#
[snarfed] a good experiment will be to follow yourself on w3c.social, or have someone else there follow you. if your next post then shows up there ok, that will be good confirmation
#
[snarfed] (and, ugh. apologies again. bad bug on my part!)
klymilark and pixelglade joined the channel
#
[tantek] I'll unfollow and refollow on w3c.social
#
[tantek] one moment
#
[tantek] oh heck, the account that was following me is no longer following me!
#
[tantek] uh [snarfed] if by "reset your followers on each remote instance" you mean caused every follower to unfollow me, then yes
jjuran joined the channel
#
[tantek] [snarfed] I think you may need to manually update Bridgy Fed's follower list for me to drop anything older than when the "deleting your actor" incident occurred
#
[tantek] hmm, spot checking and a few seem to be still around
pixelglade, GuestZero, pixelglade_, sp1ff, Guest6, barnaby, [Steve_Duncan] and [schmarty] joined the channel
#
[snarfed] yup, good call. unfortunate that Mastodon (not sure about others) evidently drops new posts on the floor if they get delivered and they're valid but there aren't any active followers
#
[snarfed] igh
#
[snarfed] er, sigh
#
aaronpk oh that's interesting
#
aaronpk it's very much a messaging protocol, not a web publishing or even data sync protocol
#
[snarfed] AP? yes, sure, but that seems unrelated...? if an instance is storing and serving a local view of a remote actor's profile, and it gets an inbox delivery of a new post (object) from that actor, I don't know that AP says anything about whether that new post should or shouldn't become visible on their profile
#
[snarfed] audience targeting, sure, but that seems separate
#
[KevinMarks] [snarfed] we webmention redirects - with the http vs https, wwwl vs bare domain and trailing / redirects being common I found it made sense to preserve anything in the redirect history
AramZS joined the channel
#
[snarfed] [KevinMarks] definitely! I'm curious about standard compliance specifically though. eg if http://target/a redirects to http://target/b , and http://source links to http://target/b , and you send a wm with source=http://source target=http://target/a , does that pass wm verification? 🤷
bterry1 joined the channel
#
aaronpk no
#
aaronpk the target parameter in the webmention request has to be the same string as appears in the page
#
[snarfed] got it! thanks
gRegor and anyx joined the channel; anyx left the channel
#
[tantek] aaronpk do we have wm rocks tests for these scenarios?
#
aaronpk For the redirect scenarios?
#
[tantek] yeah
#
[tantek] basically to have test case canonical answers for each of the questions about redirects and webmentions asked in the past 24hr here
#
[tantek] both on the positive (your implementation MUST do this)
#
[tantek] and the negative (your implementation MUST NOT do this)
hacknorris joined the channel
#
hacknorris is it better to make temporary database in excel/libreoffice or normally mysql? wanted to make a blog post reviewing hoards of things and before to sort this smh so i will be able to make sections in this post…
#
aaronpk most of the tests are for senders, we had very little tests for receivers
#
gRegor hacknorris, sounds like it would be a <table> in your blog post?
#
hacknorris no, lists, i'm gonna sort by metadata
#
hacknorris but i'm gonna sort before it going into post
#
hacknorris like a bulk review of things…
#
hacknorris like 20 to 50 apps at once each with different pros and cons…
#
gRegor Do you want the reader to be able to filter and sort them?
#
[tantek] aaronpk, I think it's also interesting that webmention receivers are I believe harder to build, so could benefit from more tests to help guide their construction
#
hacknorris gRegor : i will build base of post on that…
#
aaronpk i agree
#
aaronpk we were going for spec accuracy in the test suite initially, rather than "helpful to builders"
#
gRegor It sounds like it's basically going to be static HTML then, not sure what mysql or spreadsheet adds
#
hacknorris to sort it before lands in post
#
gRegor I guess, but seems like extra work to put it in there then copy it out into HTML vs writing the post in HTML and moving things around if you need to
#
hacknorris can you imagine a blog post with ranking of apps? that's it but i'm doing it manually and more readable as in sections
#
hacknorris and wanted sections to base on antifeatures/features
#
gRegor Sure, sort of like a top 10 blog post (or top 20 to 50) in this case :)
#
hacknorris sort of yea
#
gRegor Haven't written much like that myself, just thinking I would probably prefer to write about each of them in a document, then move them around as needed before publishing.
#
hacknorris then wanted to see all list and to manually choose to which sections add tags
#
hacknorris i want to do it before to know what exactly to write…
#
hacknorris moment. lemme count browser tabs. will say you how much exactly
#
hacknorris it's 35 apps to sort
#
hacknorris (tag, split into 5 same sections and sixth for worst, and write post)
#
hacknorris like 5 sections for 5 apps and sixth section of 10 more crappy
#
superkuh If I were doing it each thing would be a project-itemname.html fragment (no <HTML>etc header,) and combined into different pages /projectname/compilation1.html /projectname/compilation2.html, using server side includes on those HTML pages like <!--# include file="/projectname/project-itemname.html" --> <hr /> <!--# include file="/projectname/project-itemname2.html" -->
#
hacknorris i don't have serverside
#
superkuh Dang.
#
hacknorris (neocities)
#
hacknorris :)
#
[tantek] aaronpk, yes re: "spec accuracy in the test suite initially", and I think that would be a good way to focus any additional tests, per the citations of the spec you have made in answering the "redirect" questions folks have asked
#
hacknorris okay, a joke with neocities → it's codeberg without ci (aka - same features) :)
[schmarty] and vile joined the channel
#
trwnh question about u-in-reply-to vs. u-quotation-of -- am i correct in thinking that these are mostly disjoint? or are there situations where it makes sense to use both?
#
trwnh question 2: when parsing microformats, should you / does the parser dedupe links by their href? for example, in a post that is technically a reply to another post, where the post being replied to is quoted twice: i am thinking that this should be marked up as a.u-in-reply-to on both <a> elements.
#
vile here?
#
aaronpk yep
#
vile I used kiwiirc, but I liked this a lot
#
vile I used kiwiirc, but I liked this a lot
#
vile I used kiwiirc, but I liked this a lot
#
vile I used kiwiirc, but I liked this a lot
#
aaronpk it's not really an IRC client, it's mostly a viewer of the chat logs with a tiny IRC client in it
#
aaronpk also apparently it has a bug and posted your message 4 times 😂
#
vile oh hahah it is reaining hard here my internet is unstable
#
vile I am looking it in github, kudos man! I didn't know indie web had an IRC
#
aaronpk also bridged to slack and discord! https://indieweb.org/discuss
#
trwnh another question actually: i'm realizing that my "source link" is currently actually scoped to its parent h-cite and not the h-entry. is this a pattern that anyone has encountered before, or is there a better class i can use to link the h-cite to the thing it's citing? and can i specify that the purpose of the cite/quote is specifically to reply to it, or does that kind of metadata only belong on the top-level h-entry?
#
trwnh and finally a 4th only tangentially related question: the .h-cite has a child .p-author which includes a nested .h-card that has a .u-photo.u-logo with alt, and also a .p-name -- but the .h-cite .p-author is picking up this alt text on the .u-photo.u-logo... should i remove the alt text on the avatar within the h-card? this seems sub-optimal because i'd rather keep the alt text if it's possible to do so while excluding it from the .p-author
#
trwnh parsing.
#
trwnh i might be wrong and maybe the avatar is considered "decorative" in which case it should have empty alt text, but i'm mostly looking for other opinions on this
#
trwnh scratch question 3. i somehow forgot .u-url exists within h-cite
#
[tantek] trwnh, do you have an example this markup like on a post or other page to look at?
#
trwnh let me see if i can put together a paste or gist or something
#
trwnh https://gist.github.com/trwnh/020a3d2dc55d8d7dabe9fc3adb6191c8
#
trwnh ^ [tantek] i am thinking that the .u-photo[alt] should be an empty string to avoid it ending up in the mf2 parsing of .h-cite .p-author
#
[tantek] yes I'd agree and as long as it's meaning is just the same as the person's name, that makes sense
#
trwnh re: question 2 i think i was just confused, the answer is clarified by question 3 (use .u-url on the h-cite's link, which has no bearing on the outermost h-entry)
#
[tantek] interesting. I'm not sure we have had much use of e-content inside an h-cite and I'm not sure that fits
#
trwnh and in some way question 1 is also clarified by the same thing? (put the .u-in-reply-to outside the .h-cite) -- i'm just still somewhat curious about the semantics, if they are indeed disjoint (i.e. a reply and a quotation are always not the same thing?)
#
[tantek] right, a reply is more like a comment you are making on someone else's post
#
[tantek] now you can make a comment and quote part of what they said in your post, people do that sort of thing in GitHub comments all the time for example
#
trwnh so i'm thinking it should have a top-level .u-in-reply-to link but then each quotation is actually an h-cite which is unrelated and has no bearing on the metadata of the h-entry
#
[tantek] makes sense
#
trwnh also [tantek] re: the use of .e-content my understanding is that it gets parsed in a way that preserves both the html and the plaintext, so that's why i'm using it. because sometimes the html of the thing i'm quoting/citing is important and i want to preserve it (consider tags like <em> or <strong>, or the usage of multiple <p> tags within the blockquote...) -- are you saying that i should also use .p-content as well? i'm not sure
#
[tantek] or maybe e-summar?
#
[tantek] er
#
[tantek] e-summary
#
[tantek] because it's not the entirety of the contents of the thing the h-cite is linking to
#
trwnh oh hm, wasn't aware that's an option
#
[tantek] yes the prefixes and property names are kinda independent
#
gRegor I've used e-content in h-cites
#
[tantek] the prefixes are parsing instructions essentially
#
[tantek] and the names of the properties are semantics
#
gRegor what is h-cite
#
Loqi h-cite is the microformats2 vocabulary for marking up citations and references on web sites https://indieweb.org/h-cite
#
trwnh i guess the point i'm questioning is will anyone understand e-summary correctly? like the mf2 parsing algo accounts for this and the output json just has `summary` with `html` and `value`
#
trwnh if we're going off the assumption that most people use p-content and expect to directly take the value of content instead of having to take the content.value or content.html
#
gRegor consumers *should* fallback to `summary` property if there's no parsed `content`
#
trwnh i guess it'd be nice to know that e-summary even exists as an option, because this isn't mentioned or called out on various pages of indieweb.org or microformats.org
#
gRegor Yes, the parsing algorithm is generic
#
gRegor You could have `e-anything` :)
#
gRegor But summary is a common property, also used in h-entry
#
trwnh well now i have another stretch goal lol: detect if my hugo .Summary is plaintext or not, and selectively add either .p-content or .e-content (or should i always just use .e-content regardless? should i always use *both* regardless?)
#
trwnh i think using both causes duplication, so i am weakly leaning toward "always use .e-content" or possibly "do the fancy detection thing"
#
trwnh although again, some other opinions or guidance here would be really nice
#
gRegor Though if this is your example above, an h-entry with u-in-reply-to, then nested h-cites, I'd be surprised if any consumers would do anything special with those h-cites. Most would probably just take the content or summary property of the h-entry
#
gRegor Thinking specifically of the URL you're replying to, if they were parsing it to show it as a comment.
#
gRegor Using both p-content and e-content isn't needed, though now I'm curious how it would parse, haha
#
gRegor I think of it like "How might I want consumers to use this?" If I'm writing a comment with multiple lines using <p> tags, I'd use e-content in my reply. The consumer can still choose to use the `value` key if they want just a plaintext version though.
#
[tantek] I mean it's not clear any implementation does anything special consuming h-cite
pixelglade joined the channel