#dev 2022-04-21
2022-04-21 UTC
darkkirb, sp1ff, marksuth1, [Will_Monroe]1, jacky and boehs joined the channel
mro joined the channel
# 
eb Webmentions could (unperferably) be done client side, but microblogging is harder. I could run a separate server (so the blog neltify and my own microblog server) but this feels like two silos, very unsatisfying especially when I am already rolling content types (https://boehs.org/locations), other forms of categorization (https://boehs.org/tags https://boehs.org/states), and interconnecting posts of all types using [[wikilinks]]. everything is tightly nitt
mro joined the channel
# 
Murray @eb: I think that's a common roadblock/criticism of static sites. There are a few ways that people can tackle it; you've already mentioned offloading some of the work to the client-side, but that really does only need to be _some_ of the work, you can always cache things like webmentions so that they are picked up by the build next time
# 
Loqi Murray: capjamesg[d] left you a message on 2021-09-25 at 10:45am UTC: Inspired by our discussion about phone books: https://indieweb-search.jamesg.blog/results?query=jamesg.blog&serp_as_json=direct
# Murray Personally, in terms of load times, I am moving towards a model where specific parts of my site are split off into their own codebases and then interlinked with subdomains. I'm very much considering this with microblogging, which is something I don't do much off but could build independently. As you mention, though, if you want deep integration then it's perhaps not ideal
# Murray Otherwise, Netlify have just launched Edge processes. I know people are already getting tools like Remix and Astro to run _entirely_ on the Edge, which could potentially provide very dynamic, effectively personalised builds for users or groups of users. Could be something worth looking into.
omz13 joined the channel
# 
jamietanna1 eb I know folks who've got Netlify functions for their Micropub server, so it's still serverless and reduced effort, and that triggers a push to Git for their site to then rebuild. Out of interest, how long is a "slow" Netlify build for you?
gxt joined the channel
# jamietanna1 I have Hugo as my static site, then have a couple of external services to do things, but it can 100% be done through Netlify
# jamietanna1 ie https://github.com/CodeFoodPixels/netlify-plugin-webmentions to send on site deploys
mro and tetov-irc joined the channel
# 
petermolnar I've always been surprised anyone cares about that metric
# petermolnar > especially if you want to reply through bridgy etc
# petermolnar for that use case, static sites might not be the most ideal things to be honest.
# [jamesg483] petermolnar Static site build times is something I keep thinking about even though it doesn't matter too much to me.
# [jamesg483] To be honest I want to make my site a bit simpler. I realise I maybe added a bit more than I need.
# [jamesg483] My site takes about 4 minutes to build.
nertzy and mro joined the channel
[Murray] joined the channel
# petermolnar > static site builds are rarely incremental
# petermolnar you can sort of make it so, but indeed, the aggregation pages are tricky. I keep intermediate files eg. the already converted markdown to html via pandoc as cache/tmpfile so it can be picked up during generation if the content of the rest didn't change
# petermolnar these are usually the things ssgs overlook
[benjaminchait] joined the channel
# 
[benjaminchait] fwiw, my site (jekyll via netlify) takes ~4 min for a full build, but will incrementally update in ~1 min when i add a new post or update something
[Will_Monroe] joined the channel
# 
jeremycherfas is currently identifying with https://xkcd.com/1987/
# Loqi [XKCD] Python Environment https://imgs.xkcd.com/comics/python_environment_2x.png
mro, gxt and jacky joined the channel
# lagash I have plenty of Python software on my system, and the ONLY time I've ever had to deal with anything like virtualenv's was with Sagemath and heavy-duty ML stuff.
# lagash At that point you're basically doing scientific computing and should try to make everything as reproducible as possible anyways, so..
# petermolnar I stopped using envs and just go for pip install --user
# 
GWG After some debate, I decided to just publish this and let people comment and I'll edit it. https://david.shanske.com/2022/04/21/indieauth-spec-updates-2022/
# 
@dshanske IndieAuth Spec Updates 2022: https://dshanske.com/b/1Ye (twitter.com/_/status/1517168901976727553)
# jamietanna1 Yeah we discussed that - I've ended up doing it, but only to allow it to work automagically with OAuth2 libraries that support it
# 
[KevinMarks] That's valuing ease of parsing over ease of publishing
gerben joined the channel
# jamietanna1 GWG it's a good post, thanks for writing it up!
# jamietanna1 > and requires some form of authorization
# jamietanna1 I'd change that to "some form of authentication", instead
# jamietanna1 Maybe also change:
# jamietanna1 > Some clients may have been using the verification process, and should remove this.
# jamietanna1 to "and must remove this"? Cause it shouldn't be happening, and makes it clearer to folks who are doing it incorrectly
# jamietanna1 GWG++
mro joined the channel
# jamietanna1 Unlikely to be many. I believe Indigenous for Android still does it (cc marksuth) as I've seen some tokens hitting my old IndieAuth token verify endpoint
# jamietanna1 Well, not sure about "many", as I'm not sure if we're aware of how many do it?
# 
[schmarty] wellknown--
# eb jamietanna1: the actual build of my static site is 6ms per page on ci so currently max 11 seconds. The big problem is the additional overhead of ci (due to other issues I can't use conventional ci providers), so really it's 0.006x + 100. Anything above 15 seconds a build scares me, because I think it's possible I will reach the point of a few thousand posts a year (assuming 10k pages that's 1m). Moral: netlify needs to deploy faster and I need incremental s
neceve and [aciccarello] joined the channel
# 
[aciccarello] I know 11ty is working on an incremental build but progress has been slow
# [aciccarello] Right now it's basically only for reloading dev builds but the goal is to support cold start from cache
gRegor and hs0ucy joined the channel
# [aciccarello] Yeah, my fear is that it's so versatile that figuring out those dependency graphs is going to be impossible
# [aciccarello] I'd be okay with limiting features if it meant standardization and better optimization
# jamietanna1 eb interesting that you're already concerned about build times. My site takes ~5 minutes (~20k posts) to build on Hugo -> AWS S3
# jamietanna1 I think building more regularly to start with is probably OK, and then see the incremental slowdown and solve it then? Instead of pre-optimise by building once a day?
gRegor joined the channel
mro, [jeffpaul], [jeremycherfas] and jacky joined the channel
[snarfed] joined the channel
jacky, amy1, nanoflite and tetov-irc joined the channel
# 
@kevinmarks ↩️ This is pretty much how Activity Streams 1.0 worked, with WebSub for knowing when updates happened. The newer innovation is responses with webmention. (twitter.com/_/status/1517279560097742848)
# 
[tantek] yeah that tweet tree is an interesting one worth contributing to with citation on the wiki: https://twitter.com/samnewman/status/1517222858669371393
# 
@samnewman So, a silly idea, but as a tech stack for a decentralised twitter-alike, what about a federated ATOM based system? Can have disparate systems for ATOM publishing, and then pull them into a feed for your timeline. Thoughts @simonw? (twitter.com/_/status/1517222858669371393)
# [aciccarello] I thought I remember seeing the criticism on the wiki somewhere
# [aciccarello] There's a little bit here https://indieweb.org/WebFinger#Well_Known_URLs