#dev 2020-12-18

2020-12-18 UTC
[KevinMarks], [schmarty], geoffo, hirusi, alex11, yenly, schmudde and sebbu joined the channel
#
@iamhirusi Is there an existing tool/API that syndicates actual content to syndication endpoints? Or is this a limitation of http://brid.gy and not the syndicators? #indieweb #help #lazyweb (twitter.com/_/status/1339907578168393728)
#
@kkd Scrapboxにwebmentionが欲しいと書いたが設計思想に反するらしい。個人的には http://scrapbox.ioを越えたbacklinkのイメージなんだが合わんのかなぁ。https://scrapbox.io/shokai/承認欲求の刺激につながる機能を全て排除する (twitter.com/_/status/1339932456426868736)
#
@kkd Scrapboxにwebmention - tkskkd-world - Scrapbox: https://scrapbox.io/tkskkd-world/Scrapbox%E3%81%ABwebmention (twitter.com/_/status/1339937051052036098)
hirusi, gxt, geoffo, jolvera, [Ian_Forrester], [KevinMarks], [tantek], [chrisaldrich] and schmudde joined the channel
#
@m_strehl @AaronGustafson may I ask you a short question re: jekyll-webmention_io? Will `jekyll webmention` somehow determine which posts are new since last build, or does it send webmentions to all URLs of all posts, counting on idempotency on endpoints? (twitter.com/_/status/1339982958539780101)
[Raphael_Luckom] joined the channel
#
[Raphael_Luckom] Are there any good resources on average traffic capacity for _servers_ (vms, etc) based on size? Like ideally I'd want to know max requests / time handle-able by 1) NGINX serving a static site; 2) generic blog-y wordpress site. Preferably broken out by RAM and CPU
#
[Raphael_Luckom] maybe bytes / time handle-able would be easier
[snarfed] joined the channel
#
[snarfed] nginx benchmarks should be straightforward to find, maybe even from its own site
#
[snarfed] “generic blog” site will involve many more factors, probably ranges across at least a couple orders of magnitude
#
[snarfed] wordpress almost certainly has public benchmarks too though
#
[Raphael_Luckom] yeah, I can look up nginx. I think that in the other cases that "orders of magnitude" is also right. I guess I can pick a case to think about or I can try to make a generalization.
#
[snarfed] i’d expect 10qps per reasonable sized server on a reasonably optimized “generic blog,” +/- an order of magnitude
#
[snarfed] i expect nginx is more like 100-1kqps
#
[Raphael_Luckom] reasonably sized server being ~1cpu / ~5G RAM?
#
beko I don't follow. nginx has not much impact on e.g. WordPress as this is run by an fpm php process (usually) and not by nginx and this can be all over the place depending on setup.
#
beko It has however great caching out of the box :)
#
[Raphael_Luckom] right...I'm considering two different cases:
#
[Raphael_Luckom] 1. nginx serving a static site (basically trying to compare e.g. "your own nginx server" vs "github pages")
#
[Raphael_Luckom] 2. wordpress on a server that you run, where you have to handle whatever your server / lb is, your DB, and your php.
#
[Raphael_Luckom] I think #2 is still probably not quite specific enough, but I think that there are some common scenarios that we could identify
#
jacky nginx will perform excellently on those specs
#
jacky even half of that
#
[Raphael_Luckom] yeah, that's what I'm seeing on the benchmarks--nginx is pretty impressive: https://www.nginx.com/blog/testing-the-performance-of-nginx-and-nginx-plus-web-servers/ .
#
[Raphael_Luckom] It kinda seems like any self-hosting option, once it scales big enough, will have separate graphs for compute vs bandwidth. Like at the high end of those nginx graphs, the cost of compute per request is tiny, but the bandwidth would likely be similar between nginx and wordpress, and it might dwarf the compute cost.
#
[snarfed] correct, more or less
#
[snarfed] although at that level, there’s way more going on too, ie many more cost centers than just bandwidth and compute. but still, for just a simple webapp, true
#
[snarfed] eg i stored http://huffduff-video.snarfed.org/ ’s files on S3 for a long time, but bandwidth eventually grew to >$100/mo, so i switched to backblaze, which costs ~1/10th as much for bandwidth
#
[snarfed] peering agreements, whee
#
[Raphael_Luckom] I had an excessive-logging snafu that blew up my AWS bill by ~600% ($1.20 to $8.20, but still not something I'm keen to repeat) so I'm trying to be a bit more methodical
#
[snarfed] heh, yes! you may also be interested in https://snarfed.org/2020-06-27_bridgy-stats-update-5 , specifically the hosting costs graph below the fold
#
@TheRealNooshu ↩️ I use Webmentions on my static blog. Documented the process here: https://nooshu.github.io/blog/2019/11/18/implementing-webmentions/ Didn't like the perf impact of Disqus: (twitter.com/_/status/1340003831875313673)
#
[Raphael_Luckom] That graph is really interesting! I'm not sure I understand the definitions of bridgy vs backfeed vs other though
schmudde joined the channel
#
[Raphael_Luckom] nvm, they're defined in the repo
[tantek] joined the channel
#
@JulianKlode ↩️ On a related note, there also is the webmentions standard for comments in the indieweb community (twitter.com/_/status/1340010297348775939)
gRegorLove, [Ana_Rodrigues], [chrisaldrich] and [KevinMarks] joined the channel
#
@AaronGustafson ↩️ It tracks what's been sent so it doesn't resend a mention, but it will try to send mentions in older posts just in case the mentioned store has added a webmention endpoint in the interim. There's a cool off period too. (twitter.com/_/status/1340029399958134784)
[fluffy], [Rose], [jgmac1106] and [Raphael_Luckom] joined the channel
#
[Raphael_Luckom] finished writing my cost-explosion postmortem: https://raphaelluckom.com/posts/postmortem_000.html
KartikPrabhu joined the channel
#
[Raphael_Luckom] thanks for all the input, it helped me think through it all better
KartikPrabhu and [snarfed] joined the channel
#
[snarfed] [Raphael_Luckom]++
#
Loqi [Raphael_Luckom] has 2 karma in this channel over the last year (3 in all channels)
[tantek] and [KevinMarks] joined the channel
#
[KevinMarks] if you want a better answer to logging than cloudwatch, have a look at hoeneycomb.io
#
[KevinMarks] oops, honeycomb.io
#
[KevinMarks] it's a very different model where you instrument everything and look at call graphs with timing. The free tier is good to get the hang of it.
#
[jgmac1106] gwg two different flexboxes one for portrait and one for landscape is the fastest way to make a photo grid. Get any fancier and want to play on both vertical and horizontal access you need grid
#
[KevinMarks] this sounds interesting https://blog.cloudflare.com/cloudflare-pages/
#
[jgmac1106] but I would start with CSS Grid if you plan on continuing to hack on it. the link @KartikPrahbu posted is great
#
[jgmac1106] there was a bunch of posts recently about masonry layout in css grid
#
[jgmac1106] oops posted link in other channel
#
[jgmac1106] gwg just start with the most basic Grid layout and let it grow with you, warts and all, Grid is so logical it can turn engineers into designers
#
superkuh As long as it's okay that browser older than 3 years don't support your website at all.