#dev 2021-09-08

2021-09-08 UTC
#
[tantek]
seemed like an interesting dev read on things that we've debated discussed about personal sites and plumbing before: https://www.baldurbjarnason.com/2021/single-page-app-morality-play/
#
[tantek]
snarfed, re: bluesky contest
#
[tantek]
s/IndieAuth is a official W3C standard (ie Technical Recommendation)./IndieAuth was published as a W3C Note (https://www.w3.org/TR/indieauth) by the Social Web Working Group, and is subsequently maintained as a Living Standard by the IndieWeb community (https://indieauth.spec.indieweb.org/)
#
[tantek]
other than that, looks great!
#
[tantek]
thumbs up! and happy to be noted as a reviewer if you wish. great writing! snarfed++
#
Loqi
snarfed has 27 karma in this channel over the last year (56 in all channels)
#
@kevinmarks
↩️ Depends which bits of the twitter experience you want; you have a blog, you could integrate webmentions and hook up http://brid.gy to get twitter replies. You could use websub to read feeds and post responses with micropub
(twitter.com/_/status/1435412624540377097)
#
@kevinmarks
↩️ Depends which bits of the twitter experience you want; you have a blog, you could integrate webmentions and hook up http://brid.gy to get twitter replies. You could use websub to read feeds and post responses with micropub
(twitter.com/_/status/1435412624540377097)
mdas and KartikPrabhu joined the channel
#
[snarfed]
thanks [tantek]! much appreciated
[fluffy], maxwelljoslyn[d] and jonnybarnes joined the channel
#
capjamesg[d]
Looks like you need to email them snarfed. There is an email on the Satellite page.
jonnybarnes joined the channel
#
capjamesg[d]
[chrisaldrich] re: PESOS for feeds I like that idea. I'd need to have a use case for myself first but I like the idea 😄
hendursa1, jonnybarnes, tetov-irc and [timothy_chambe] joined the channel
#
@NeilKNet
You want to cross-post between Twitter and Flickr and then 15 years later https://twitter.com/blaine/status/1435317481393074192
(twitter.com/_/status/1435319982326378499)
hendursaga, jonnybarnes and hhg joined the channel
#
[snarfed]
aaronpk i fixed that granary twitter list bug, it should now show the right URL in the UI and support both list slugs and ids ok
hendursaga and angelo joined the channel
#
aaronpk
oh yay thanks [snarfed]++
#
Loqi
[snarfed] has 28 karma in this channel over the last year (57 in all channels)
#
Loqi
😊
KartikPrabhu and [aciccarello] joined the channel
#
GWG
I need someone familiar with IndieAuth to help review WordPress IndieAuth PRs. Is anyone available? Even if it is just a PHP person looking...
jamietanna joined the channel
#
jamietanna
GWG share the links - I can have a look at some point, but if others are around to help too please do!
#
GWG
jamietanna: This is for the start of Token Introspection https://github.com/indieweb/wordpress-indieauth/pull/217
#
Loqi
[dshanske] #217 Introspection
#
GWG
This adds the option to turn a create scope into a draft scope even if not originally requested https://github.com/indieweb/wordpress-indieauth/pull/213
#
Loqi
[dshanske] #213 Turn create into a create/draft/none option selection
#
GWG
And this adds an experimental Ticket Auth receiving endpoint https://github.com/indieweb/wordpress-indieauth/pull/212
#
Loqi
[dshanske] #212 Ticket Endpoint
#
GWG
If I get these looked at, next is refresh tokens. Then I will be up to date with the spec and current pending items
#
GWG
jamietanna: Also wrote the PR proposal for refresh https://github.com/indieweb/indieauth/pull/90
#
Loqi
[dshanske] #90 Refresh Tokens
#
Loqi
[dshanske] #94 Token Introspection
#
GWG
And accessing protected resources https://github.com/indieweb/indieauth/pull/95
#
Loqi
[dshanske] #95 Add section on accessing protected resources
Ruxton, bitmega, shoesNsocks and strugee joined the channel
#
[tantek]
are the results really undesirable aaronpk? I think we should be ok with "thin" results like that if they're not above more relevant results
#
capjamesg[d]
How should I distinguish category / emoji / tags / whatever from a regular post feed?
#
aaronpk
capjamesg[d]: what I was thinking as a way to ignore those aggregate pages is something like if the post contains several h-entry objects and each has a URL property then ignore those pages
#
[tantek]
not sure that's worth it?
#
[tantek]
instead, prefer "more relevant content" and the others will get pushed down
#
aaronpk
i personally don't like seeing those in search results especially because it's likely that the content that was indexed at that page may now be gone from the page
#
aaronpk
s/ignore/downrank if you want
#
[tantek]
nah, I disagree with "ignoring" the pages. much better to just have them lower down in the results
#
capjamesg[d]
[tankek] this is sort of a problem for more “competitive” queries.
#
capjamesg[d]
Where sometimes lists rank high because they have a lot of relevant heading / mentions of a keyword.
#
[tantek]
I'd say look for the dt-published and prefer more recently published results
#
aaronpk
but especially for my /all page, or even my home page, the content of that page is replaced every couple days or more
#
capjamesg[d]
A category of pages tagged as Coffee on my blog will contain lots of Coffee pages.
#
aaronpk
so i've already found a case of clicking on something from that search engine not actually having the result on the page
#
[tantek]
if a page/post doesn't have a dt-published (or dt-updated for that matter), then treat it as if it were published in 1970-01-01 😛
#
capjamesg[d]
and will probably rank well for them too because of the density.
bitmega joined the channel
#
capjamesg[d]
Yeah. Feed pages are subject to change.
#
capjamesg[d]
And I haven’t set up a regular reindex yet.
#
aaronpk
it's a lot more work for the indexer, but personally I would rather never see feed pages in the results, and instead the indexer would go crawl the individual pages from the feed page and only return those
#
capjamesg[d]
But that’s already coded. I just need to instruct the crawler to recrawl when I am ready.
#
capjamesg[d]
aaronpk love it.
#
aaronpk
basically the way i would expect an "indieweb search engine" to work is that it is searching posts rather than web pages
#
capjamesg[d]
There is a 1000 page crawl budget so I’d much prefer to let sites spend it on important content.
#
capjamesg[d]
I could basically treat them as follow, no index.
#
capjamesg[d]
And index links instead.
#
capjamesg[d]
(Pages linked to from the category page, sorry)
#
capjamesg[d]
I agree re: searching web pages. I have been using the engine mainly to find docs / examples / code / fun articles.
#
capjamesg[d]
[tantek] I have thought about recency.
#
capjamesg[d]
It’s going on my list of ideas.
#
capjamesg[d]
Agree that recency is good especially for some terms that are time bound.
#
capjamesg[d]
What was the example of a result without the term on page aaronpk?
#
Loqi
It looks like we don't have a page for "example of a result without the term on page aaronpk" yet. Would you like to create it? (Or just say "example of a result without the term on page aaronpk is ____", a sentence describing the term)
#
capjamesg[d]
No thanks loqi.
#
aaronpk
"habanero" returns my /all page
#
capjamesg[d]
I’ll check the index tomorrow.
#
capjamesg[d]
Good to know.
#
capjamesg[d]
I can weigh feeds of old content lower in rankings. Not sure if I’ll remove them though re: [tantek]s comment.
#
aaronpk
on the "slightly odd but technically correct" end of the spectrum, "habanero" also returns my https://aaronparecki.com/emoji/%F0%9F%93%BA page where there are a bunch of unrelated posts on top of it
#
capjamesg[d]
Interesting.
bitmega joined the channel
#
aaronpk
but that's a good example of why i am thinking of this as a "post search engine" rather than traditional. i'd expect only the one post on that page to be returned in the "habanero" search
#
[tantek]
yeah, given that you have a crawl "budget" capjames[d], you may want to noindex pages that lack a dt-published or dt-updated
#
[tantek]
maybe with an exception for home pages? since those are often relevant regardless of the search term used?
bitmega joined the channel
#
[snarfed]
one difficulty is that you'd probably only want to hide a feed page if you _know_ that all of its entries are also in the index and will be served as individual results
#
[snarfed]
which you could do, just takes logic
bitmega joined the channel
#
aaronpk
i like what capjamesg[d] said earlier of treating feed pages as a list of links to follow rather than index the pages themselves
#
capjamesg[d]
aaronpk particularly curious to hear from you on queries that have a broader surface area / are likely to be addressed by multiple sites.
#
capjamesg[d]
The engine was designed only for my blog so it has almost been a full rewrite.
#
capjamesg[d]
I have most of the logic now for integrating with the microformats / IndieWeb wikis but I need to start looking at relevance for them.
#
aaronpk
"recipe" returns a nice variety of sites!
#
capjamesg[d]
Maybe a bit too heavy on mine but I write a lot about recipes
#
capjamesg[d]
That’s sort of where tag pages are helpful haha 🙂
#
capjamesg[d]
Because a few show up on that list that would not if they were no indexed.
#
capjamesg[d]
OAuth is almost all your site 😂
#
capjamesg[d]
I really need to fix those meta descriptions.
bitmega and tetov-irc joined the channel
#
[snarfed]
npd re symmetric linking, the main discussion in the contest entry was IndieAuth, which only works if you can log into (or otherwise prove ownership of) one of the linked accounts
#
[snarfed]
ie you could rel-me link to any twitter account, but you couldn't log into it unless it's yours
#
npd[m]
ah, I honestly wasn't thinking as much about IndieAuth at all
#
npd[m]
like I thought the Mastodon example is a really strong one, and maybe not as highlighted as it could be. Mastodon can verify links (and provide some evidence along with a checkmark) without an authentication step, just using the publicly observable symmetric rel-me links
#
npd[m]
when I was looking at this, it actually seemed to me like the main challenge is that a surprising number of accounts don't allow for user-controlled profile fields to point back to a personal URL
#
npd[m]
like the Matrix spec includes profile information, but it's not really implemented much yet
Seirdy joined the channel
#
[snarfed]
right. hence we don't strongly depend on rel-me links inside silos
#
[snarfed]
added a link to XFN!
#
npd[m]
yeah, I think it'll help in the future to encourage that. many people will encounter me through some silo'd service originally (like this Matrix account I'm chatting with right now), and we'd like to be able to provide verification to them of my other accounts
#
npd[m]
my other thought was that I'm not sure you need to fill out the rubric yourself, that's just how Jay et al. will be evaluating proposals
#
npd[m]
and maybe more to the point, I don't think this would be scored low on originality -- we're evaluating the tradeoffs of using this existing technology to apply to this use case
#
[snarfed]
oh yeah I know it's their rubric, just figured I'd address the criteria directly
#
npd[m]
okay cool
#
[snarfed]
heh sure, we'd happily take a 3 for originality. just wanted to call out that long-lived building blocks are positive, as opposed to novelty
#
npd[m]
I think the symmetric links are more effective than service-by-service authentication in order to show common identity
#
npd[m]
there are caveats about repeatedly testing verification over time, and making sure the links are provided the same to everyone, but I think the robustness comes from having those links public and easily auditable by anyone
#
npd[m]
and if we wanted to talk about areas to make improvements, I think it's making it as easy as possible to add links in any new service's profile, and ideally to specify that those links are that person with rel-me
#
[snarfed]
bidirectional rel-me vs IndieAuth serve different use cases, and both are useful
#
[snarfed]
the more important part is that almost none of the big silos support them. github, mastodon...and i've run out, at least of any others I know offhand
#
[snarfed]
...support bidirectional rel-me)
#
[snarfed]
we're obviously all for getting more of the silos to support them! we just don't have much weight to make that happen
#
npd[m]
yeah, I think the stop-gap is to interpret profile links from known silos as implicitly identity-providing. so that the symmetric link is still some kind of verification, even though ideally it would be symmetric with both sides explicitly stating identity
#
npd[m]
that's what this XFN page describes under "Identity consolidation": http://gmpg.org/xfn/and/
#
[snarfed]
...but you then have to know where in the silo profile the "my web site" field is. ie on a twitter profile, you wouldn't want to accept any link in their recent tweets. which implies scraping, and that way lie dragons.
#
[snarfed]
ooorrrr you can use an API, which is only sometimes available, and may require approval etc
#
npd[m]
but your suggested alternative is to implement OAuth dances for every such silo, right? that seems like a much higher burden of API implementation
#
[snarfed]
right, neither of those is great. rel-me is ideal but almost no one supports it
#
npd[m]
but yeah, figuring out implicit symmetric links for large silo's requires some custom code for those silos, to grab profile data and to determine that some links or fields are useful for inferring identity
#
[snarfed]
yup. and that custom code will regularly break with no warning or easy way to fix it other than wading through their HTML and avoiding their bot filtering again
#
[snarfed]
if you're concluding, gah, everything has problems, nothing Just Works...welcome we're glad you're here! 😆
#
[tantek]
minor correction to that. everything real/practical/deployed/in-use has problems, the only things that "just work" (in people's minds) are aspirations and theoretical proposals. see also Nirvana Fallacy.