2024-03-04 UTC
[tantek] joined the channel
# 00:02 [tantek] I have never used (posted) URLs with userinfo "@" and am very close to cutting that out of my regex to simplify it # 00:08 angelo tantek can you give an example of a @-url-with-path ? [KevinMarks] joined the channel
# 00:08 [KevinMarks] I haven't touched that regex for a bit (I copied it into the Python version a while back) but are you close to the point where an explicit list of tlds is clearer than the compressed 2 letter combos? # 00:16 [tantek] angelo, @-url-with-path is for folks who want to use @ for their silo profile URL, e.g. @instagram.com/tantek # 00:19 [tantek] users do type and publish @-mentions @-@-mentions and @-url-path mentions # 00:22 angelo what would an example @-username-@-url-with-path be? # 00:22 [tantek] WebFinger-- is a very limiting model from this perspective # 00:22 Loqi WebFinger has -8 karma in this channel over the last year (-9 in all channels) # 00:22 [tantek] angelo, services that have some path segment between the domain and username # 00:23 [tantek] or even someone that installed a Mastodon server inside a directory instead of at root # 00:38 angelo ( (@{user|domain})?@ | https?:// ) {domain|uri} --- instead of --- (@{domain}) | https?://{domain/uri} # 00:39 angelo there would be a false positive with @-domain-@-url-with-path # 00:43 [tantek] what's false about it? also some systems allow usernames with dots in them which then sorta resemble domains # 00:45 aaronpk well it's either @user@domain or @domain/path but there's no way @user@domain/path is possible in activitypub right? # 00:48 Loqi basicauth has -1 karma over the last year # 00:48 [tantek] the user:password portion is called "userinfo" and that's exactly what I'm talking about dropping # 00:48 aaronpk i wouldn't be surprised if it's dropped from the URI spec in a few years at this point # 00:48 [tantek] it currently *does* work in my autolinker but I've never used it, nor do I know anyone who has ever used it # 00:49 angelo aaronpk i believe something along the lines of @tantek@flickr.com/photos # 00:49 [tantek] it really ought to be discouraged, so I'm going to make that policy decision in this update # 00:49 angelo tantek you're right about the dots being fine # 00:50 Loqi urluserinfo has -1 karma over the last year # 00:50 Loqi basicauth has -2 karma over the last year # 00:50 aaronpk now i want to see someone try to spin up an activitypub profile where @user@domain/path is the only correct way to @-mention them # 00:50 Loqi webfinger has -9 karma in this channel over the last year (-10 in all channels) # 00:50 aaronpk i wonder if mastodon or even anything would be able to work with that # 00:51 [tantek] aaronpk, install Mastodon inside a directory instead of root and find out! # 00:51 real_devastatia After reading all of this, I'm actually glad filter_var excludes anything that isn't a personal website URL from my address book. # 00:51 [Joe_Crawford] Hasn't u:p in the front of a url been deprecated for like 20 years? # 00:51 [tantek] because presumably that's how you'd address that instance inside a directory # 00:51 [tantek] "Note: _username_ & _password_ were omitted for simplification (and you shouldn't be using http-auth anyway); _params_ omitted because it's obsolete." # 00:52 aaronpk [Joe_Crawford]: no it's definitely still part of URLs, and browsers changed from visiting the page immediately to showing a prompt first, but it still workss # 00:52 [Joe_Crawford] !!! I didn't know that. I've not tried it in a long time. Probably with FTP. # 00:53 [Joe_Crawford] ```URI producers should not provide a URI that contains a username or # 00:53 [Joe_Crawford] password that is intended to be secret. URIs are frequently # 00:53 [Joe_Crawford] displayed by browsers, stored in clear text bookmarks, and logged by # 00:53 [Joe_Crawford] user agent history and intermediary applications (proxies). A # 00:53 [Joe_Crawford] password appearing within the userinfo component is deprecated and # 00:53 [Joe_Crawford] should be considered an error (or simply ignored) except in those # 00:53 [Joe_Crawford] rare cases where the 'password' parameter is intended to be public.``` # 00:53 [Joe_Crawford] January 2005. I'm stunned to know it's extant in active specs. # 00:54 [Joe_Crawford] "should not" but still. [Joe_Crawford] joined the channel
AramZS joined the channel
# 01:33 [tantek] ok regex refactoring and update done to match what I want to match, and it got shorter 🙂 # 01:34 [tantek] (thanks to dropping the deprecated userinfo matching bits) # 01:36 [tantek] now the matching code update since I simplified what gets matched to make all the combinations of @ and @-@ "work" burley, gRegor, gRegorLove_, burley_ and jacky joined the channel
# 03:51 [tantek] got it working. In addition to more thorough @-@ support, @-domain now also supports domains with hyphens (previous version broke), and @-@-s with domains with non-ASCII characters in them like https://loðfíll.is/@briansuda burley, bterry, box464, [Al_Abut], jacky, [dominik], gRegorLove_, gRegor, [jeremycherfas], [Paul_Robert_Ll], [Murray], [lcs], [kimberlyhirsh], Guest6, AramZS and to2ds joined the channel
# 15:07 to2ds 🤯 When you finallly discover link header is a "thing" and not just there for decoration. [schmarty] joined the channel
# 15:17 [tantek] Just realized that Slack (iOS mobile app at least) auto-links @-@-mentions too! (Like the briansuda example above) # 15:21 [tantek] Here's what it (still) looks like in Slack (on iOS mobile, a different client and device than I wrote it!) burley joined the channel
[aaronpk] and burley joined the channel
# 15:58 [tantek] I mean it's consistent with how Slack does other auto-linking like if I type a plain domain, slack links it and the logs see a fully qualified URL. I was just surprised that they had @-@ autolinking built in! jacky, burley, mahboubine, sivoais and gRegor joined the channel
# 17:47 gRegor I think the last one I got from BF was a follow on 3/2 # 17:51 gRegor [tantek], yeah there was a confusing moment in the logs yesterday where you were talking about autolinking not working for "tantek.com" but it showed up as "http://tantek.com". I was confused until I remembered Slack adds the scheme :) # 17:59 [tantek] indeed and this is now my use-case for "@" for any scheme-less URL # 18:34 [tantek] btw if anyone sees any linking "weirdness" on my home page or post permalinks, please LMK ASAP! # 18:34 [tantek] I *think* everything is working with my autolinker update (zero regressions so far, and all new features functioning as expected), however I realize it's always easy to miss something burley and jjuran joined the channel
# 19:55 gRegor snarfed, webmentions for my posts today are coming in though, so maybe a temporary glitch [snarfed] joined the channel
barnaby, gRegorLove_, burley, Renfield, bret and jacky joined the channel
[KevinMarks] I haven't touched that regex for a bit (I copied it into the Python version a while back) but are you close to the point where an explicit list of tlds is clearer than the compressed 2 letter combos?
[tantek] for reference: https://en.wikipedia.org/wiki/URL#Syntax (where "userinfo" is described literally)
real_devastatia After reading all of this, I'm actually glad filter_var excludes anything that isn't a personal website URL from my address book.
aaronpk speaking of which, when do we get an update to this? https://tantek.com/2011/238/b1/many-ways-slice-url-name-pieces
gRegor snarfed, a couple Mastodon replies might not have sent a webmention (nothing in my received log at least). Al's reply today and Joe's a couple days ago. https://techhub.social/@alabut/112036502516191474