GWG_Fiddling with the code I use to retrieve sites for Post Kinds and Yarns. Trying to figure out how to determine if a page is a feed. Comment is that my current method of fetching the entire page is slow.
[Rose]Ok, if I were to start writing my own MicroSub server, and a reader, which one would you start with? I feel like this is a chicken and egg problem. (To be clear, these will be separate projects)
sgregerHi. Since the Wiki seems down (?) allow me a simple question: how should I deal with incoming webmentions where the source page does not have a h-entry?
ZegnatBut for the general ecosystem health, building a server would be much appreciated. There are multiple readers now (Monocle, Together, Indigenous) but very few readers used (from what I hear Yarns is sometimes a bit unstable and I do not know if anyone uses Ekster?)
sgregerI recall (and interpret the spec that way) that MF are not per se a requirement for a valid webmention. But how should I deal with it? Process it as a Pingback? (Without the MF metadata, it literally is not more than that...)
ZegnatYou recall correctly, sgreger :) Webmentions are content independent, they just express a link relation between source and target. According to the spec, you only need to verify that the link relation exists and then the webmention is valid.
ZegnatWhether you want to try and transform them into a comment to display is out of scope for webmentions themselves. You could try falling back to JSON-LD or anything else if there are no microformats on a page
sgregerZegnat: Thanks :) So all I need to figure out is how to process (and subsequently display) that. I like the JSON-LD fallback idea. Otherwise I think something along the lines of showing it with webmentions that are not replies (e.g. "This post got mentioned on example.com")
[Rose]And set up a Symfony demo app, because my boss said if he had to choose between Symfony and Laravel for work he thinks from what he's read he would like Symfony more, so time to try a project with it!
ZegnatI like Symfony/Zend more than Laravel because of its modularity. Also might be biased as https://github.com/Nyholm/psr7 is available as Symfony Flex package :P
[Rose]Which may not be true, but it definitely feels like it if someone can't walk you through the code base (someone can also be excellent documentation)
[Rose]That's also possible, but if we want people to install this stuff breaking the 'back end' of feeds up into multiple parts will not make things easier for the user who just wants to install something, rather harder.
[Rose]Honestly watchtower was not easy for me to set up, so I would like to incorporate that into the main microsub server to make it easier for the end user who just wants to install it on their server
ZegnatI find it almost more important that I can limit the amount of results than what format the results come in. Want to keep load as low as possible so it is more feasible to handle multiple sites.
ZegnatIf you, for some unknown reason, want to add yourself to indiewebsearchengineproofofconcept the _links->self->href needs to refer to a JSON version of the post, which in its turn has to contain title->rendered, content->rendered and link keys…
[Rose]I had a crazy idea on the polling, which was that when your reader calls your microsub server it goes out and fetches stuff. But that's not a great idea for all sorts of ideas.
ZegnatThat is why you want some sort of asynchronous thing running in the background that keeps your feeds up to date. And suddenly you are back to Watchtower ;)
[tantek]So the problem with microblog UX for likes in an Atom feed (that have fallback content) I think is demonstrating the problem of Atom’s lack of extensibility in practice to publish new types/kinds of posts & interactions
[tantek]We had a big session on this at IWS 2018 with Manton and Aaronpk and I don’t think we actually captured (or maybe followed up on) all the conclusions like how to handle p-summary, what to do with properties you don’t recognize etc
sknebelI know it was discussed afterwards too, with questions about things like "how do I tell if a property is for something important that needs fallback shown or not", and it didn't seem straightforward to solve
[tantek]No we solved it at the IWS 2018 session but it required updating a bunch of things (specs, how to sections) that I don’t think we finished doing
[tantek]Anything that’s beyond a note, article, reply, because all h-entry consuming code is expected to support those since there so many of those out there
ZegnatAnd your likes have the p-summary, and that also seems to be what shows up on micro.blog. I think the general thought is that the fallback value is not actually a good enough fallback?
Zegnat[tantek]: you can click links on micro.blog/t. Which actually means it isn’t using p-summary, because that wouldn’t contain the links. As a non-micro.blog user I can’t speak for their app though. I only checked your public profile page
[tantek]Am I the only one who cares/worries about “owning my labelings” or do others want to own the act of adding (or removing) a label (essentially a category/tag/hashtag) on an /issue post, e.g. on GitHub?
sknebelI regularly have problems with twitter links on mobile, where it loves to claim you're "rate limited" even if you haven't interacted with twitter in days, but not login prompts
ZegnatI just tested the first tweet linked on micro.blog/t, and it covers the tweet with a transparent grey box (dimming the page, leading to less contrast / harder to read) and puts a banner in view urging me to download their app
sknebel[Rose]: IMHO the major limitation with shared hosting is needing some kind of cron. if you've got that I don't think its unreasonable, and there's possible hacks around that too (but those are hacks, so effort/unreliable/...)
sknebel[Rose]: right, using some external service or event to trigger regularly is one of the hacks. I know free website monitoring services have also been popular for that: they consistently send you a request every X minutes
ZegnatI think PHP can spawn a subprocess itself as well. If you want the main thread to already send a response to the browser while you do sort of a fake-cron triggered by people visiting the site
sknebel"I imagine most people with serious interest will end up with a VPS anyway" - I'd not count on that. E.g. how many serious users of e.g. WP are on (fancy) hosting instead of a VPS because they can "get away" with that? (some of that is of course specialized WP hosting, which is different again, but not all)
aaronpkone of the reasons I use it is because I know they've done a good job of isolating all the user accounts and making sure things can't eat up all the resources
petermolnarthere's no target url in the source, and this is true. My previous attempt to fix this was to add u-syndication from the source to flickr, but it didn't do anything. Any ideas?
aaronpkah yeah i had that problem too, bridgy couldn't find my replies since they aren't on my home page, so I made an /all page with everything and bridgy is polling that
snarfed[tantek]: if it helps, for me the difference in your like UX between your site and atom feed is that your site very effectively collapses them all into (seemingly) a single post, whereas micro.blog and feed readers don't
snarfedi wouldn't mind if they similarly showed up in my feed reader collapsed in a single item. i definitely mind when they're 30-50 individual items.
[Rose], gRegorLove__ and [frank] joined the channel
[frank][Zegnat] re open source RSS readers, I'd say RSSOwl (but it's written in Java...) and Tiny Tiny RSS. The last one has quite some features. I haven't used it in a while but I know [tonz] has a self-installed version.
snarfed[tantek]: hah, no. that's an interesting idea, but aggressive, and could be error prone. "similar" isn't necessarily easy to get right. i'm not sure i'd want that feature, at least not globally
snarfedi understand the idea and motivation, but individual sites have way more semantic knowledge of which of their items can and should be collapsed. readers etc don't, so it'd be significantly harder for them to do it confidently
ZegnatI am more likely to want my reader to expand it (show the actual thing that was liked) than to collapge it. But I guess that is also preference on how to consume things.
[eddie]aaronpk I do think at some point we'll either need a way to communicate filters like that to the Microsub server through Microsub OR have some type of session exchange where you can open up your Microsub server's settings url already signed in
[eddie]You really think that Microsub should have to define all the possible filters and if a server wants to provide additional functionality you should have to go log into the server instead of your reader app?
aaronpkin aperture you have the option to either include *only* specific post types, include posts matching specific keywords, or exclude specific post types or exclude posts matching keywords. that's a lot of options and i'm not even sure everyone actually wants all of them
aaronpki don't think it makes sense to expose *all* those settings in a reader app, but then if only some of them are exposed that also seems weird to split the configuration up
[eddie]I agree, not ALL of those should have to be included but having a weird splitting of settings provides a "wait, where do I have to go for that?"
Loqi[sknebel] A compromise might be a way to discover the link into the backend admin UI. Then the reader could send the user directly to the right channel/feed settings instead of the user having to manually navigate the backend UI.
[eddie]The thing I'm thinking of experimenting with UI wise that I would like to do is based on tantek's general person-first approach. So you choose a person to follow (likely a firehose feed), by default they go in the standard "Global" feed, but if you want you could click to "split between channels", which would in the backend essentially subscribe to that feed from two different channels with different post types included
[eddie]So from the UI you are subscribing once, but I can then say "I want to follow aaronpk, and all his posts should go here, but all likes and reposts would go over there, and ignore everything else"
[tantek]The other thing to do with likes/reposts is to use them to create a synthetic "discovery" feed of all the things all your friends are liking/reposting
[tantek]so instead of seeing each individual like from your friends all liking the same thing, you see the thing *once* in the discovery channel, perhaps with a mini facepile or something of all your friends that liked it
LoqiIt looks like we don't have a page for "suggested setup for splitting one feed into different channels" yet. Would you like to create it? (Or just say "suggested setup for splitting one feed into different channels is ____", a sentence describing the term)
aaronpkthis idea of "magic" or "smart" channels has come up a few times for various purposes, and I like the idea, so I'm thinking we may need to add a flag that tells the app that a channel can't be deleted or modified because it's being controlled by the server. in that case, you could do something like have only one main channel that people add stuff to, and then the server can decide to make posts appear
aaronpki guess what i'm saying is you should feel free to make channels do whatever you want them to do, and add any UI necessary to your own microsub server
ZegnatTurns out there are a million different ways to do feeds ;) E.g. for discovery I thought Fever had an interesting system were you would add certain firehose feeds only as kindle/sparks (or whatever they were called) and those would be used to surface things from other feeds.
[eddie]In regards to the earlier discussion about "Collapsing" common post types into a single post (like tantek's likes) I've created an issue for Remark to keep in the back of my mind and investigate while I'm working out the UI
LoqiRemark is a native Social Reader app for iOS that is replacing some functionality previously available in Indigenous for iOS https://indieweb.org/Remark
petermolnarok. today's summary of "I had 15 minutes here and there and did stuff on my site": konami code css, opensearch definition, brid.gy flickr webmention backfeed seems to be fixed. Useful day.