#dev 2025-02-27
2025-02-27 UTC
# 
immibis a URL in reality (and also in theory!) specifies how to get a resource. it's not a unique identifier for a resource. a URN is an identifier for a resource and is more rarely used.
# immibis a URI is either a URL or a URN. all http:// URIs are URLs, unless someone is abusing them, as many protocols do.
# immibis (such as XML namespaces)
# immibis an authentication tag is part of how you fetch a resource, so it can be part of a URL
# immibis it's also possible for ambient authorization to be implied
# immibis trwnh: ^
aciccarello joined the channel
# sebbu immibis, but most browsers and libs removed the old http://user:pass@domain.tld/path/file url format !
# immibis yeah, some kind of security concern. I think that you could spoof the user part to look like a whole URL. Doesn't mean it's not logically a valid URL though.
# immibis iirc it IS valid in the username and password field - that's why it was a problem. Otherwise it would be no different from http://google.com.reallylongstring.evil.com/
newbster, GWG and grufwub joined the channel
# trwnh immibis: you wouldn't generally embed an oauth token into a uri, would you? no, you would use the Authorization header
# trwnh i also don't think the URL vs URN split makes sense anymore. rather, they are all names, it's just that some names encode protocols for dereferencing or resolving the name
# 
tantek.com yeah, urns were a dead-end to begin with.
# tantek.com that was my attempt to kill the topic here. not sure any indieweb site bothers with any URN URI or auth-in-old-URL syntax stuff. likely better for #random unless someone is actually talking about building support for it on their personal site
# immibis trwnh: have you heard of an unlisted video (or other object)?
# immibis URLs vs URNs definitely make sense...
# immibis the only reason you wouldn't put a username and password in a URL to curl is that authorization systems got more complex.
# tantek.com ^ let us know how you're building these things on your personal site, otherwise there's no empirical basis to claim "definitely make sense".
# tantek.com "reason you wouldn't put a username and password in a URL" is it's bad security. please see the actual discussion in the specification on this instead of making specious claims like "authorization systems got more complex". https://github.com/whatwg/fetch/pull/465
# tantek.com [edit] "reason you wouldn't put a username and password in a URL" is it's bad security. please see the actual discussion in the specification on this instead of making specious claims like "authorization systems got more complex". https://github.com/whatwg/fetch/pull/465
Boiler, oodani, leo60228, vriska, [Jo], [morganm], IWSlackGateway, [aaronpk], vskate and [KevinMarks] joined the channel
# 
capjamesg[d] [Joe_Crawford] I would love to learn about fit-content at a future FrESH.
# capjamesg[d] I can't join the next one as it conflicts with HWC Edinburgh though.
nemonical and [artlung] joined the channel
vskate, Guest6, ttybitnik, [pfefferle] and [Ana_R] joined the channel
# immibis @tantek.com "bad security" has nothing to do with whether something is a valid conceptual part of a URL or not
[qubyte] joined the channel
# 
[KevinMarks] https://bsky.app/profile/middlesbroughfe.co.uk/post/3lj5un72wdk2d Ana is already speaking there
axcelott, GuestZero, nemonical, [schmarty], [tantek] and [Jo] joined the channel
# capjamesg[d] [KevinMarks] Is there a more up to date version of https://github.com/kevinmarks/cassis-autolink-py
# capjamesg[d] [edit] [KevinMarks] Is there a more up to date version of https://github.com/kevinmarks/cassis-autolink-py
# [KevinMarks] No, I haven't ported Tantek's improvements since. He's definitely updated the regex since then
# [KevinMarks] Kyle copied over the existing tests iirc.
# [KevinMarks] The strategy is manual until it hurts
# [KevinMarks] We could look at the mf parsing model where we define input and output in a language independent format maybe
dhrjarun joined the channel
# capjamesg[d] I can PR an update with the regex?
# capjamesg[d] I tried jamesg.blog in the autolinker and it failed.
# [KevinMarks] How does cassis run tests in both languages?
sp1ff and gRegor joined the channel
# capjamesg[d] Starbucks' RSS feed is blocked by their Cloudflare 😭
bterry and nemonical joined the channel
# capjamesg[d] Starbucks is using Cloudflare and blocks my reader's requests to its feeds 😭
# capjamesg[d] The URL is https://about.starbucks.com/feed/
# capjamesg[d] A HTTP request asks the user agent to try again with cookies and JS enabled.
# capjamesg[d] With various mentions of "cf" (cloudflare) in there.
ttybitnik joined the channel
# [tantek] does it also block a http://granary.io request for the feed?
# capjamesg[d] Granary gets a 403.
gRegorLove_, Zegnat and [snarfed] joined the channel
GWG and bterry joined the channel
# [snarfed] I'm above my pay grade in CSS right now. the bluesky icon on my posts on https://snarfed.org/ is a ::before pseudo-selector with url(...) for an image and display: inline-flex width: 16px, and it works great in Firefox, and degrades gracefully in Safari, but is totally broken in Chrome
# catgirlin.space uh.. is it supposed to be that big? lol (Safari Technology Preview release 213)
# 
gRegor [snarfed], This worked with my fiddling in Chrome: https://stackoverflow.com/a/8978010
[benji] joined the channel
bterry, sebbu2 and sp1ff joined the channel