#dev 2024-03-18

2024-03-18 UTC
[Joe_Crawford] joined the channel
#
[Joe_Crawford]
capjamesg looking at cp_snowfort.webp It seems to be 826x506 pixels which I can imagine on a 2x mobile screen will appear slightly more pixely
geoffo, [snarfed] and [Fran] joined the channel
#
capjamesg
[Joe_Crawford] Do I need a bigger image?
#
[Joe_Crawford]
I think it couldn’t hurt. Though since it’s a background you don’t get to use the nice image size choice algorithms available with srcset, at least that I know of.
ttybitnik, jacky, shoesNsocks, [Scout] and [aciccarello] joined the channel
#
[aciccarello]
Reposting [Scout]'s question:
#
[aciccarello]
"i want to set up a site for my mother's birthday party where people can upload photos. like a mini instagram of sorts. and i want to give a level of protection by adding a password so not every person/bot on the internet can a) see my mum's baby pics and b) upload stuff to my site. so that rules out github pages for hosting. i am trying to decide where is a good place to host it. i know netlifly can handle both the password and the
#
[aciccarello]
dynamic uploads. but if i understand this right, it would mean that the site gets rebuilt every time someone posts a picture. i expect the vast majority of use/uploads to be happening on the birthday party evening. and i about the 15 minute built time limit. am i worried for no reason? because if there are more than 100 uploads that night, i would be surprised. on the other hand, i would very much like to avoid any surprises if possible
#
[aciccarello]
and know what i am getting myself into...
#
[aciccarello]
any thoughts/suggestions/ideas?"
#
[aciccarello]
I'm wondering if you want something long lasting or a page that's only up for a couple months or less.
#
[snarfed]
letting users upload pictures will be somewhat involved, far beyond github pages
#
[snarfed]
honestly the easiest place to start might be a shared album on eg Google Photos or iCloud etc
#
[snarfed]
or maybe the equivalent photo app on a self-hosted cloud service like owncloud or yunohost
#
[snarfed]
unrelated, [jacky] re https://shkspr.mobi/blog/2024/03/well-known-avatar/ , I'm trying to understand the benefit of this over a rel=avatar link in webfinger
#
Loqi
[preview] [Terence Eden’s Blog] .well-known/avatar https://shkspr.mobi/blog/2024/03/well-known-avatar/ Hot on the heels of a post I wrote 4 years ago, wouldn't it be useful to have a well-known URl for user avatar images? When I sign up to a web service, I don't want to faff arou...
#
[snarfed]
the only benefit I see is that it saves one HTTP request and then pulling that link out of the returned JSON
#
jacky
[snarfed]: they're going to say it doesn't need a HTML/XML parser (from that post)
#
jacky
which _shrugs_ is kinda fair but then you also got a JSON parser if you're doing conneg for webfinger
#
[snarfed]
just JSON, not HTML/XML
#
[snarfed]
if the only motivation is to avoid parsing JSON, that seems kinda crazy and unlikely
#
[snarfed]
I must be missing something
#
jacky
hence the Accept: image/* path
#
jacky
I think it could be for _hyper-simple_ tooling (like doing it from a terminal or something that's meant to be "dumb", like linking this directly in the browser)
#
[snarfed]
maybe, but you'd think he would have said that
#
[snarfed]
I feel like we're over-speculating
#
jacky
Oh yeah, I am lol
#
[aciccarello]
Yeah, the avatars suggestion seems like they want to inline a well-known URL into an img tag.
#
jacky
best option would be to reply to this post tbh
#
jacky
[aciccarello]: yup!
#
[aciccarello]
I did see that they specifically mentioned h-card.
#
[aciccarello]
[Scout] Do you need to update the website immediately or could you publish the photos after the fact?
#
[aciccarello]
Something like a Google Form would let you collect photos from friends and publish them later.
#
[Scout]
[snarfed] i was going to host the photos elsewhere. i was more concerned about the constant re-building of the site every time someone posts a photo. but maybe i am misunderstanding the meaning of built time?
#
[Scout]
[aciccarello] yes the idea was that everyone could upload photos as the evening progresses and the site would be projected on the wall and also i was gonna print little qr codes for people to scan so they can open the page on their phones. that sort of thing. i don't want to curate a nice album at the end, i wanted it more like when people post on insta from an gig/event. but personalised to one person/family. i have thought of airtable. but
#
[Scout]
also s3 bucket. because it would be good practice for my personal site too.
#
[aciccarello]
I wouldn't worry too much about build times. Netlify gives you something like 300 minutes of build times credits and only runs one at a time so that should be fine for an event.
#
[snarfed]
[Scout] if you really want to build this yourself, ok! but yeah since arbitrary external users are uploading content, it doesn't sound like a good fit for a static site
#
[aciccarello]
I agree, I'd probably default to a dynamic site to make it easy to control auth and uploads.
#
[aciccarello]
If you wanted to later, you could archive it to a static site later to reduce hosting costs.
[Murray] joined the channel
#
[Murray]
If you use something like Astro, you'd be able to build with server rendering then switch it to static at some point later on, and keep everything else the same 😉
#
[tantek]
Obligatory wellknown--
#
Loqi
wellknown has -1 karma over the last year
#
aaronpk
i'm a big fan of static sites for archiving that kind of thing but yes i would not build it as a static site first
#
aaronpk
for a specific event, use something with server side code, but find/build something that you know will be easy to flatten to a static site after
#
aaronpk
because you also don't want the maintenance burden of keeping the dynamic code of the event site online years after the event is over
#
[tantek]
The "requires a server to parse HTML" is a bs excuse. They're all parsing HTML already to do link previews, unfurling etc
#
[tantek]
Parse for representative h-card, use the photo properly. Done. No personally website? Fallback to gravatar. Yes it's centralized, however at least it's automattic
#
[tantek]
Use the photo property*
#
aaronpk
yeah everyone needs to get over the fear of parsing HTML
#
aaronpk
the only legitimate argument against parsing html is for embedded devices like esp32 microcontrollers. for those I agree adding an HTML parsing stack is too much overhead
#
redblobgames
Parsing html makes me angry so I avoid it ;-)
#
aaronpk
parsing arbitrary HTML for arbitrary structures isn't trivial, but that's often not what is actually being talked about in these kinds of discussions
#
[tantek]
And often in these kinds of discussions, the potential customers (servers that setup new accounts) are already parsing HTML
#
[aciccarello]
Yeah, I would expect you need fallbacks in 90% of cases anyway so you'll need to parse something.
#
[aciccarello]
Also, what about different image sizes?
[contact898], [schmarty], gRegor, [campegg] and [KevinMarks] joined the channel
#
[KevinMarks]
Don't we have an existing pattern of using the link rel=icon for that?
[qubyte] joined the channel
#
[qubyte]
Obligatory _parsing headers is hard too!_ (in can you’re thinking of putting that link in a header) ;)
#
capjamesg
Has anyone done any work with knowledge graphs?
#
capjamesg
I have been thinking a lot this evening about how LMMs could be used for data retrieval from a document to make a knowledge graph that can then be queried.
[lcs] joined the channel
#
[lcs]
Isn’t that RAG?
#
[lcs]
(Retrieval-Augmented Generation)
#
[lcs]
My knowledge of it is cursory, but that was my understanding of the concept
#
[snarfed]
sounds related but a bit different? capjamesg you're thinking of a first LLM pre-processing step that generates a traditional knowledge graph, maybe directly from embeddings, and then later querying that knowledge graph directly?
#
[snarfed]
maybe => #indieweb-chat
bret joined the channel
#
GWG
Can someone tell me why someone added the Apache license as a comment to my PR? https://github.com/indieweb/indieauth/commit/9a294ace6ea2a80b72f712ca16cda17277ac7157#commitcomment-139927353
#
[tantek]
GWG looks like a spam test
jamey joined the channel
#
Loqi
[preview] [BARCOLAIRTHAILAND] #471 Jk BMS home as
#
[tantek]
I went to their profile and clicked the "Report" link and reported them for spamming an org I'm a part of. We'll see if GitHub does anything
#
GWG
interesting
ttybitnik joined the channel
#
[tantek]
I have seen enough nonsense like this on GitHub that I have basically zero patience left for it
#
GWG
I was just excited someone commented on my IndieAuth PR until I... wasn't
jamey_ and [tantek] joined the channel
#
jacky
TIL that GNUnet is making moves
[capjamesg] and [Ros] joined the channel
#
[tantek]
only "informational" RFC but still