#GWGZegnat, what do you think about multiple rel values?
#ZegnatI like them. Some URL can express being author as well as me simultaneously. Or both your indieauth and token endpoint simultaneously. Etc.
#ZegnatI do not know how I feel about treating multiple rel values as having a singular meaning when they stand together. But I think that is specced as a possibility?
#GWGDoesn't rel="home alternate" mean it's the home page and it's an alternate independently? (as asked [1], [2], [3])
#GWG No. That's not how rel works. Rel values combine. E.g. rel="alternate stylesheet" has worked this way for over a decade. Rel values also combine with other attributes, e.g. use rel="alternate" media="handheld" on a link to a separate mobile-specific version of a page. Tantek 20:56, 10 August 2011 (UTC)
#ZegnatFeels to me like rel-home introduces an extra case to the alternate in addition to what is already speced in the HTML spec. Seems fine to me.
#ZegnatI think rel-feed is nice in addition to rel-alternate as rel-alternate is specced to only refer to a feed together with 2 specific type-attributes. While rel-feed gets to widen this to any document type, e.g. JSON.
#ZegnatSo seems like you are pretty much spot on in all your descriptions in your blog post, GWG
#ZegnatAnything specific you still feel unclear about? Or would you just like more eyes on the discovery algo you penned?
#ZegnatAlthough I do not see any practical reason why we would care
#Zegnat<link rel="stylesheet" href="style.css"><link rel="alternate" href="style.css"> makes very little sense.
#GWGI only use rel, so I need to switch to rel-urls
#[tantek]I think that "same element" semantic is specific *only* to "stylesheet", so yeah that's true. However there's no use-case anyone's found to worry about that from an mf2 perspective
#ZegnatBut theoretically that should mean the first link is to the main stylesheet, the second to an alternative representation of the current document. It does not mean style.css is an alternative stylesheet. But that latter is what one would think when looking at the mf2 output
#Zegnat[tantek]: I agree with that, like I said, I see no practical case where this could go wrong.
#[tantek]we could perhaps add a warning about the "stylesheet" value in particular, but I don't expect anyone will do any stylesheet reasoning using an mf2 parser. that would be quite odd
#Zegnat<link rel="home" href="a.html"><link rel="alternate" href="a.html"> - would surely have the same limition, in that per HTML spec this is defining 2 unique “links”. Neither of them claiming that a.html is a feed for the homepage. But the mf2 output would suggest that it is.
#ZegnatNonsensical example again, because the sheer semantics of it seem super wrong :P
#[tantek]alternate does not mean feed all by itself
#ZegnatAnd I do not know any rel values other than alternate that this is important for, so ... meh? Seems fine to use rel-urls
#[tantek]that's the bigger problem with your reasoning
#ZegnatYes, from mf2 it looks like feed for the homepage, because mf2 might think there was a link to a.html with the rels "home alternate" on it at once
#ZegnatAnd that is defined as a feed for the homepage
#ZegnatOh, hmm, or I am reading rel-home wrong and that one actually *also* requires a type
#GWGSo should more people do rel-home alternate on feeds?
#ZegnatAh, it seems like I was mistaken and it also requires type. Woops. So for the mf2 output to be wrong(-ish) you need even more convoluted examples that are even less likely to go wrong in the real world!
#Zegnat<link href="a" rel="home"><link href="a" rel="alternate"><link href="a" rel="ignorable" type="application/atom+xml"> - Never declares a feed, but the mf2 output might suggest that a is in fact a feed of the homepage
#ZegnatBut even I find this to be too much of an edge-case :P
#ZegnatGWG: for now a small tweak to your algo is that whenever rel-alternate is considered, you also need to consider the type value. Only together with a feed type is it considered to mean feed.
#ZegnatAccording to the WHATWG spec it is only a feed when the type is either application/rss+xml or application/atom+xml.
#ZegnatYeah, I am not sure it is a good idea to assume <rel="alternate" type="application/json"> can always be assumed to point at something following the JSONFeed spec.
#ZegnatSounds logical? Although someone should define that rel-feed changes the meaning of rel-tag / rel-author in that case
#ZegnatI do not think datetime is a valid attribute on links though
#ZegnatAnd then there may be a period where people get confused with parsing. E.g. before they have understood that rel-feed changes the meaning of rel-author they may believe a page to have 2 distinct authors: <link rel="author" href="me"><link rel="feed author" href="my-feed"> = 2 authors me and my-feed.
#GWGI just want to be able to identify the type of feed
#ZegnatAlso note that `tag`, with its current definition, is not allowed as a rel on <link> elements. So probably not going to be used for feed links up in the HTML head
#GWGZegnat, data attributes are allowed on link elements
#GWGI do want to use home and feed to allow feed discovery though
#ZegnatAccording to the HTML standard it is fine to have a lot of rel-alternate + type <link> elements. They are “not necessarily syndicating exactly the same content as the current page”.
#ZegnatAnd then making clear which feed is the home feed by adding the home rel value makes sense to me
#Zegnatactually doesn’t like rel-tag because of its concept of a “tag space”
[Michael_Beckwit, [Lauro], KartikPrabhu, dougbeal|mb1 and [KevinMarks] joined the channel
#[KevinMarks]Well, for a classic site url which is mydomain.me/tag/indie that is a reasonable rel-tag url, but linking to it would imply tagging the homepage with that tag. If that page is a h-feed of posts with that tag, feed tag almost makes sense.
#[KevinMarks]If you're linking to it from a tagged post that's semantically OK.
#GWG[KevinMarks]: Marking up feeds is what I'm debating.
ChanServ, KartikPrabhu, mauz555, gRegorLove and [Michael_Beckwit joined the channel