[tantek][snarfed] more BF weirdness. mentioning fed.bridgy from my latest post via https://mention.tech/ seems to result in a red X (some form of rejection?) and not showing up on my BF dashboard
Loqi[preview] [Tantek Çelik] @snarfed.org posted a great overview of thoughtful (and sometimes heated) discussions across blogs and the #fediverse about how freely should “public” posts & comments on the web flow across sites:
“Moderate people, not code” (https://snarfe...
[tantek][snarfed] BridgyFed is broken both for the new post federation case, and for updating a prior post, both of which I did in the past 30ish min, sent several webmentions for, verified that yes their permalinks have links to http://fed.brid.gy inside, and see zero evidence of anything happening on https://fed.brid.gy/web/tantek.com
[snarfed]the last few times we looked at this, BF hadn't received some or many of the wms in question. sounds like there was more confidence that they were sent here, but still, just want to confirm before I dive into anything
[snarfed]and re your post's wm showing up on my post, I'm not sure about a delay, but for the current rendering as a facepile, agreed that since it's a mention, ideally it should link to your post, not your homepage
[tantek][snarfed] it looks like something was busted in BridgyFed's side because the webmentions did get sent to all the other links in the post including your blog
[snarfed]the relative times on BF dashboards, eg "a minute ago" like you saw at 6:21p, are last updated times. they get updated as the post is delivered to new inboxes. BF only lightly parallelizes those deliveries, and for >300 like yours, and receiving servers often taking a while to respond, it can easily take 20+m total
[tantek]if you want to add another column next to the green checkbox + number and yellow ! + number that indicates the time delivery ranges, that would be different
[tantek]that way we could also see how much lag there is between Bridgy receiving the Webmention and federating it out to at least the first server/follower
[snarfed]I see that timing in logs and metrics very often. it depends on the number of inboxes to deliver to, but in large cases like yours, I think it's order of 30s or so
[snarfed]I get that this was frustrating. I can definitely improve BF's dashboard UX. this is the third or fourth time that the root cause of your site not federating was webmentions not getting sent though, so in the future, maybe tone down the "Bridgy Fed is broken this way and that way" language?
[tantek]it was not clear that was the case, it looked more like Bridgy's webmention endpoint was just not responding at all since every other link was being sent a webmention
[snarfed]hopefully there are low hanging fruit fixes or UX tweaks [KevinMarks] can make there before you decide to rebuild it yourself, but I'll let you all determine that
[tantek][KevinMarks] do you keep logs of when mention.tech displays a red X so you can look at the conditions that caused that? because "link not in source" is obviously false since it found it later
[0x3b0b]I was pondering whether mention.tech might have somehow failed to fetch the entire source, but BF was one of the first relevant links in the source, not the last.
[tantek]0x3b0b exactly, the bridgyfed link is before the content of the source (but still in the h-entry), which is totally fine since reply-contexts also work like that (in-reply-to links are BEFORE the content, but still inside the h-entry)
[tantek]ok the https://webmention.app/check link explicitly says "Keep in mind that this test page only scans the 10 most recent `h-entry` elements on the target." however there is only ONE h-entry there, so that caveat should be meaningless
[tantek]same problem with that page, only " 10 webmention supported links found. " and again, did not find http://fed.brid.gy which was *before* any of the links it found in the source
[tantek][KevinMarks] mention.tech specifically https://mention.tech/mentionall seems to only be looking for links inside the "content" property which is incorrect. it should be looking for links inside the h-entry in general, since there are plenty of use-cases for that e.g. in-reply-to URLs go outside the "content"
[snarfed]apart from the ops trouble, [tantek] thank you for the kind words and the response! it's been nice to see that people have been open to more nuance and thought and angles on the topic
[tantek]apologies for the frustrated expressions [snarfed], it's been a bad 72 hours of tech failures and I realize that's not your or BridgyFed's fault
[tantek]honestly given the "I don't know if it worked or not" experience that I still feel happens too often, it would be great if the time column in the dashboard showed time of Webmention received to provide the fastest possible feedback that yes, *something* worked
[tantek]and then I understand that doing all the discovery, going through instances to deliver etc. will take some indeterminate amount of time from seconds to a few minutes
[tantek]at least BridgyFed already has that indicator of the counts of in-progress vs delivered (green check) vs some issue (yellow !) indicator so you have some idea how far it has gotten and how far it has to go
[tantek]and yes on the subject matter of our posts, I feel there are A LOT more questions than answers at this point, and it's more important to keep asking these questions rather than trying to force an answer by reasoning from too little information (which it feels many perspectives are attempting to assert, and then fighting over their conclusions)
[0x3b0b]I've got a tentative item in my posts-to-write list for sorting through my thoughts on some of those questions; not sure when or if, though. It's interesting to try to keep up with the conversation though.
[0x3b0b]By the way, [snarfed] : I believe the docs make it clear #nobridge is already implemented; has been for a while, I think. Am I correct in thinking your ideas about preferred bridge detection and bridge deduping are still on the drawing board?
[0x3b0b]I may chime in on that later - I'm definitely interested in what you come up with. I'm actually more interested on a _personal_ level in another topic that I think I remember coming up before, which is what would happen if someone pointed BF at a website they wanted to follow...but that website was actually an AP profile.
[0x3b0b]I'm also curious whether it's possible to check whether BF has a profile for a website yet without triggering it to make one if the answer is no.
[snarfed]for the first q, BF only bridges top-level web sites (domains), not URLs with paths. some top level web sites are also AP profiles, eg the instance actors for Lemmy and Friendica, but otherwise it's rare
[snarfed]sounds similar to https://github.com/snarfed/bridgy-fed/issues/348 . BF would have to detect that a web site is also an AP actor, either by fetching it with conneg, or by looking for a rel=alternate link with the AP content type
[0x3b0b]I may follow up on that sometime soon, too, for the microblogpub case. I've run into that before from the other side - I deliberately marked up a note as an Indieweb reply to one of Tantek's posts, in the spirit of elevating Indieweb domains above @-@ addresses, in addition to posting it as a reply to its BF-erated copy so that it would thread correctly in that context. So of course he got multiple webmentions.
[tantek]you might still be able to make it work with "only a link" but that wasn't the intent of the feature. your original post needs the rel=syndication link
Loqiu-syndication is a way to discoverably link from your original posts to syndicated copies on other sites like social media silos https://indieweb.org/u-syndication
[campegg][snarfed] I don’t want to pile on the BF support train, but was wondering if you had a chance to take a look at what was going on with my follows, etc? (https://fed.brid.gy/web/campegg.com) I’m pretty sure it’s something I’ve done, but don’t know how to remediate. There really is no rush; I appreciate the time and effort you put into BF!
sebbu2, tPoltergeist, barnaby and shoesNsocks joined the channel
[snarfed]we didn't start out like that, but we saw quickly that people link to posts much more often on GitHub in non-POSSEd issues, and backfeeding those issues' comments confused people, so we special cased GH to require synd links
[snarfed][campegg] thanks for the kind words! it looks like you tried to follow my web site http://snarfed.org? BF only handles cross-network follows. it figures you'd just use a normal reader to follow a web site, you wouldn't need BF at all
[campegg]Right, got it. Thanks! Secondary issue is that posts to my site don't seems to be federating, and I'm not able to follow @cam@campegg.com… I (think I) followed the 'getting started' instructions, but nothing seems to be showing up
savemeI was wondering where people host their videos and photos for their webpage? I want a dead-easy solution, but always overengineer it. (Currently syncing a local folder with vids to an s3 bucket and then using the api to get the links and loop through them in a gallery)
[tantek]Right. My point is that video storage is hard enough of an indieweb problem to push a primarily video publishing effort from IndieWeb to YouTube-only, even just for the storage aspect (since you'd still get YouTube engagement if you POSSEd)
aaronpkit would, because views from "regular viewers" mean a lot more to the youtube algorithm especially in the first few hours of publishing a video. so if my most dedicated viewers were watching on my site, it would hurt the performance overall
aaronpki actually had it set up to PESOS the video descriptions to /videos but it embedded the youtube player on the page. I stopped doing that last year tho because I decided to move it all to a different domain entirely
btremMoving from #indieweb per request, I'm trying to find some up-to-date resources for SVG and accessibility. What I've found so far is hopelessly out of date.
[tantek]I suppose I would flip it around. Start with a marked up list of percents and text labels and then style that into a bar chart visually with CSS
barnabythat’s pretty much what we do at work IIRC. either way, IMO the goal is not so much “how do I label this specific bar to make it accessible” but rather (assuming you’re targeting screenreaders) “how would this entire chart be best presented via a screen reader”
btremWell, the text is there, but the amount, not the percent. So each bar has a number, 9, or 3, or whatever. And the bar shows what percent of the whole it is. Working (more or less) version: https://btrem.com/feed/wordle
btremAnd the question is, *where* do I put the percent? What element should it be in? Another text element? And then how do I accessibly label it? There's no alt attribute for <svg/>, much less for a <rect/> inside an <svg/>.
[tantek]In the old days of SVG (20y ago), one of the "visions" for SVG was to replace HTML, which was both ill-conceived and arrogant. that's how we ended up with things like that
btremPerhaps you're right, the problem is inherent to svg. But I think part of the problem is aria, which also feels like an attempt to replace what we have with something brand new and shiny. Something to fix all the problems of the old system. Cue xkcd standards.
btremWell, re: svg, my goal is to see if I can replicate the New York Times' wordle stats page visually, but improve the accessibility. I've gotten part way there, just using posh instead of the div soup the NYT uses. So maybe I'll just need to learn to live with how it is now.
[Murray]@btrem I can see that MDN is being a bit strict around the way they're referencing aria-label, but it should largely work. I think the more "correct" way to do this is using aria-describedby and the `<desc>` SVG element. Adrian Roselli has some thoughts here: https://adrianroselli.com/2022/04/accessible-description-exposure.html
btremre: alistapart I read that years ago. I have mixed feelings about the whole sordid affair, but I'd rather not get bogged down in arguments. I'm not strongly one way or the other, just skeptical of some of the author's assumptions (which were quite common at the time, iirc).
[Murray]@btrem: if you're aiming for accessibility, one thing to consider is that your current implementation breaks a "golden rule": the accessible content is different to what I can see. Unless I'm missing something, the data visualisation only shows me what number of games were won on each successive number of turns. But the accessible label gives me a percentage. That feels confusing. If the visual label is "8", then the accessible label should
btrembut kudos for this clever remark re: xhtml 2: "It would be inaccurate to say that it was going nowhere fast. It was going nowhere very, very slowly." :-D
btremThe visual label for the *bar* is the percentage. The adjacent text has the number. Maybe I'm thinking about this wrong, but bar charts usually don't show the percentage. That's what the bars are for. But they often include the number.
btremIdeally, I'd like for the number and bar to represent the data. But to have the percent and number for those with sight disabilities of some kind.
[Murray]The confusion is that I wouldn't be able to look at that bar and infer 44%. All I can infer is that 8 is larger than the other numbers, and that this is the most common turn to win on as a result.
[Murray]The problem with adding an invisible additional label is that the ADOM now results in this: "3, 44%, 8". That doesn't mean anything. The visual inference you're really making here is "8 games were won on the 3rd turn". So perhaps that would be a closer accessible label to the visual. But personally, I wouldn't even do that. Visual labels should match accessible labels 1:1, as many sighted users still use assistive technology, and a mismatch
btremYeah, I wasn't sure how it would work to have both the number and the percentage on the bar. I thought it might look cluttered and be a bit confusing.
[Murray]E.g. if I were browsing that chart with a screen reader, as someone who is (largely) sighted, I would get to the 44% and think I'd been jumped somewhere else on the page. It isn't immediately obvious where I am any more
[Murray]Personally, I'd find knowing the percentage kind of interesting, and it wouldn't feel that cluttered, but that's a design choice. Right now, 44% is not something that I can infer from the design as a sighted person either, so if you want that information to be surfaced, I'd suggest a visible label of some kind
[Murray]One other way of handling this is a toggle or other interactive event. You could hover the bars to reveal an extra percentage label (possibly to the right of the bars, to avoid cluttering the interface on smaller ones). Or you could have a switch-like toggle somewhere to flip between "number of turns" and "percentage of turns". At least, that's the kind of suggestion I'd raise at work 😄
btremLeft filed analogy, but back in the day, there was a form to email script in Matt's script archive that was notoriously buggy, especially regarding security. And a group called London Perl Mongers (I think that was the name) created a drop-in replacement. There were a few additional features, but they mostly constrained the project to drop-in replacement. This does what Matt's does, but avoids the bugs.
btremSo I sometimes like to take that philosophy. Here, with the NYT wordle page. I did a similar thing with Open Table's awful restaurant widget. I was working for a restaurant, and also did their website, and wound up designing a far more accessible widget. And my goal was to make it look exactly like theirs, but with real html elements.
btremAlso, if I'm going to make wordle more accessible, the biggest hurdle is the tiles, which are emojis and therefore explain nothing. That's something I definitely want to tackle. Probably turn them into svg icons with a <title> something like, "letter not in word|letter in word but in wrong place|letter in right place".
e-snail, tPoltergeist and omz13 joined the channel
btremWow, I just had a serious brain malfunction. I was about to ask where I could find the emojis used for wordle results because I want to know the hex colors. And it would have to be on the web so I could use Firefox eyedropper. And I searched the web and came up empty. And, just as I was tying the question here, I realized: oh, right. On *my* site. Where I'm now posting wordle results, including the tile grids. :-/
btremBut I still have a question: in some windows, e.g., VS Code editor, the wordle tiles have a white swish in the upper left corner, almost like a shadow. On others, e.g, Firefox, they don't. Any idea why? Or how I could see a version with the swish in a context where I could try to replicate it in svg?
btremAh, so probably doesn't make sense to worry about the swish, since different users will see different things no matter what I do. Like I see it differently in Firefox vs. VS Code on the same OS (Ubuntu).
[KevinMarks]Likely they're using different emoji fonts. You can pick one you like the look of for consistency, though with squares it's a bit moot. I had this problem with stuff for BBC as I was using emoji in ui, but some of them were running very old windows versions where the emoji were black and white.
btremWell, I'm not entirely sold on emojis. Using them in my own-my-wordle project has made me a bit skeptical. (Though, to be fair, some of the problems appears to be the result of javascript text-encoding ignorance. Maybe.)
[tantek][Murray]++ thanks for that link, that was the sort of abbr[title] empirical information I was looking for. appreciated it. it always did feel imperfect to me so nice to see something to cite to explain the issues.
[tantek]that being said, I stand by my original assertion that for common symbolic (rather than in-prose) use of emojis in particular, it is quite possible to use the <abbr> element ... with an explicit 'aria-label' attribute of what you want a screen reader to say, semantically, instead of the often very verbose description of the emoji.
[tantek]there is no harm AFAIK in *also* putting that text on the title attribute, and that way someone using a device with a pointer could also hover over the element in case they were confused, or if they had trouble seeing the difference between say the yellow and green squares in a low lighting context. e.g. <abbr aria-label="letter in right place" title="letter in right place">:large_green_square:</abbr>
[tantek]span would give the same effect, however abbr feels like it is somewhat better communicating the semantic that yes, this green square emoji means more than just "green square", it is a shorthand for "letter in the right place"
[Murray]hmm, I'd want to test that in a few popular combos just to check it didn't result in duplication in places. If `title` is completely ignore than yes, that makes sense to me too
[tantek]another use-case for this is the use of emojis for list item markers, where you can use <abbr> to similar express what the emoji marker in the context of the list item means
[tantek]in my recent post using emojis for list item markers, I realize I actually mixed those two. two of the list items use emojis purely decoratively (no semantic would be lost if replaced with •), and two of the emojis represent a label of a sort for the list item: https://tantek.com/2024/022/t1/indiewebcamp-brighton-planned
sknebel(emoji have also a bunch of fun failure cases, e.g. text-styled fallbacks without color, design differences, .. (e.g. there's the colored book emoji, which on some OSes have numbers on the label and on others not...))
[tantek] [snarfed] more BF weirdness. mentioning fed.bridgy from my latest post via https://mention.tech/ seems to result in a red X (some form of rejection?) and not showing up on my BF dashboard
Loqi [preview] [Tantek Çelik] @snarfed.org posted a great overview of thoughtful (and sometimes heated) discussions across blogs and the #fediverse about how freely should “public” posts & comments on the web flow across sites:
“Moderate people, not code” (https://snarfe...
jacky might be interesting for multiple authors to a blog (or syncing micropub servers between one another) https://www.loro.dev/blog/loro-richtext
[KevinMarks] I think red X means couldn't find the target in the page source
[snarfed] the relative times on BF dashboards, eg "a minute ago" like you saw at 6:21p, are last updated times. they get updated as the post is delivered to new inboxes. BF only lightly parallelizes those deliveries, and for >300 like yours, and receiving servers often taking a while to respond, it can easily take 20+m total
[0x3b0b] I may chime in on that later - I'm definitely interested in what you come up with. I'm actually more interested on a _personal_ level in another topic that I think I remember coming up before, which is what would happen if someone pointed BF at a website they wanted to follow...but that website was actually an AP profile.
aaronpk i have the activitypub conference talk on my site tho, and all the HLS segments are saved in the git repo of my site storage https://aaronparecki.com/2020/09/22/25/activitypub-oauth-2-1
[Murray] @btrem I can see that MDN is being a bit strict around the way they're referencing aria-label, but it should largely work. I think the more "correct" way to do this is using aria-describedby and the `<desc>` SVG element. Adrian Roselli has some thoughts here: https://adrianroselli.com/2022/04/accessible-description-exposure.html