aaronpkspeaking of autolinking, i've seen a lot of youtube comments accidentally autolink text because people don't include a space after the period at the end of a sentence
[tantek]this is why it's a bad idea to base your design of UI things (hyperlinks are UI) on any sense of plumbing (someone's official list of TLDs somewhere) completionism--
gRegorGotcha. Yeah, indielogin should end up responding with `me=thebackupbox.net/~epoch` after you auth, but MW probably has an issue with the slash. No biggie, though.
[tantek]gRegor, that's odd re: chat logs auto-linking the period "." at end. Pretty sure I coded auto_link in CASSIS to specifically omit trailing punctuation like that
barnaby, AramZS, gRegorLove_, geoffo and kandr3s[m] joined the channel
[manton]@barnaby Right. I guess it’s not as clever or obvious if .new catches on… But I’m not sure .new _can_ catch on because by definition there will be some few good ones.
barnabythe recent unexpected usefulness of QR codes has made me a little more open to accepting weird gimmicks… but it doesn’t cost $400 to make a QR code
Seirdyis it worth making a webmention receiver listen over unencrypted HTTP to redirect to HTTPS, instead of just listening over HTTPS and leaving port 80 alone?
Seirdy[manton]: i'd stay away from google domains unless you use a separate google account just for buying domains. you don't want to get locked out of domain renewal because the algorithm thought your email activity was suspicious.
[schmarty]redirects from HTTP => HTTPS are more for user agents like browsers where someone might not know exactly which protocol to use. if you advertise an HTTPS url and a webmention sender uses HTTP, that's their error. if you advertise an HTTP url but don't accept webmentions there, that's your error. 😄
Seirdyalso curious if anyone has a system that handles dead links by replacing them with archived snapshots, and how that fits into webmentions and microformats
Seirdybut if you send a menchie to a "normal" (for lack of a better term) site, and later the page disappears, you might replace your link with an archived snapshot--either a self-hosted snapshot (perhaps powered by archivebox) or a third-party service (e.g. wayback machine)
[snarfed]people have also thought this through a bit for caching profile pictures, eg for displaying webmentions. haven't found that on the wiki much yet, but I think aaronpk's p3k does
Seirdysince the displayed URL will change but internally will be linked to the original id of the page. so on your backend, a page could be represented with two urls: the "original" url and the "currently live" url. the currently live url could be a pointer to the current canonical or the archived snapshot.
Seirdy[snarfed]: i'd just use max(document_cache-control_duration,7d) to cache a u-photo URL value, and use max(image_cache-control_duration,14d) to cache the actual image file. and if the former disappears, i could switch the image to grayscale.
Seirdyi'd probably post-process the image file too. shrink, zopflipng, next-gen image formats, a binary value set to "true" if the measured brightness is too high (so i can reduce the opacity for the image in my dark theme), nsfw-filter, possible manual review if it's visually distinct enough, etc. so i would probably want the min cache duration to be kinda long, since this isn't a dance i want to
Seirdy[snarfed]: well it's something i just thought of in like 5s. though after thinking about it for slightly more than 5s, i guess they might have actually wanted to remove their u-photo so i should fall back to a placeholder
Seirdywait hang on this could actually work. if you use your own domain for libravatar (as opposed to a subdomain or different domain altogether) it won't be an additional DNS lookup; it'll be just like serving a plain old image. but the URL will be machine-discoverable from just your email address at the same domain as well. so you'll get the benefits of an indieweb u-photo *and* the benefits of