#dev 2025-03-28

2025-03-28 UTC
[KevinMarks], sebbu2, box464, barnaby, alephalpha0, SylvatiCodes, rrix, doesnm, bterry1 and NaomiAmethyst8 joined the channel
#
Kolev
I've been posting directly to Mastodon instead of POSSEing. 🙁
glacier and doesnm joined the channel
#
Zegnat
pfych: do you know if you are using AWS WAF? I wonder if your experience with Cloudflare blocking users and CloudFront not doing so is because Flare has WAF by default and Front does not
#
Zegnat
for a second wanted to abbreviate all instances of Cloudflare and CloudFront as CF and make that sentence completely unparsable
#
aaronpk
i can already barely keep track of which is which, so wouldn’t make much of a difference
#
Zegnat
Next.js << [https://eduardoboucas.com/posts/2025-03-25-you-should-know-this-before-choosing-nextjs/ You should know this before choosing Next.js] - about the handling of a 9.1 out of 10 severity CVE, lock-in, and more
#
Loqi
ok, I added "[https://eduardoboucas.com/posts/2025-03-25-you-should-know-this-before-choosing-nextjs/ You should know this before choosing Next.js] - about the handling of a 9.1 out of 10 severity CVE, lock-in, and more" to the "See Also" section of /Next.js https://indieweb.org/wiki/index.php?diff=100611&oldid=88300
#
Zegnat
aaronpk: I wonder if your experience with CF blocking users and CF not doing so is because CF has WAF by default and CF does not (good luck! :D)
doesnm joined the channel; glacier left the channel
#
pfych
Zegnat: Likely! 😅
doesnm and [Murray] joined the channel
#
[Murray]
because I use Netlify, I think I am using some elements of Cloudflare, but if it were up to me I wouldn't, mainly because it annoys me how much I have to click through their irritating captcha screens 😄
#
pfych
I saw xe’s proof of work bot protection thing and I’m tempted to use it for my high traffic (by bots) pages like my gitea instance
#
Xe
hi
#
pfych
Oh damn hi
#
pfych
sorry for the ping!
#
Xe
oh don't worry
#
Xe
you're not cursing me for being a broken GPU driver, s' all good
#
pfych
I’ve got a new GPU to install tomorrow and I’m hoping that I don’t have to tinker too much with linux to get it working (Manifesting)
#
Xe
is it intel?
#
Xe
because for some reason i'm quite aware that the intel GPU driver is kinda terrible
#
Xe
:P
#
pfych
9070xt - it looks like I just need to setup testing repos and install latest mesa stuff
#
Xe
oh you got the good one
#
pfych
I am looking at intel for my NAS and I’m kinda dreading it from what I’ve been reading haha
#
Xe
i mean, it works until it doesn't
#
Xe
but that's par for the course with GPUs
#
pfych
good to know lol
#
pfych
My partners old 2070ti is currently in the NAS but it’s overkill and actually doesn’t fit in the case at all 😭
#
Xe
but yes part of the main raison d'etre for anubis atm is to protect the small web
#
sknebel
pfych: I just got a 9070xt and its been working great for now, but I also picked a distro optimizing for that kind of thing (Bazzite)
#
sknebel
(well, and I need to figure out why despite distro-patches removing Mozillas overrides Firefox wont do GPU-decoding, but thats probably not the drivers fault)
#
doesnm
indieauth-openid is not working now? Displays "invalid response type: code"
[schmarty], nemonical, ttybitnik, shoesNsocks and [manton] joined the channel
#
[manton]
[snarfed] It occurred to me while thinking about lexicons… What does Bridgy Fed do with 500-char Mastodon posts when bridging to Bluesky?
#
Xe
it truncates them and links to upstream
#
[manton]
Got it, thanks!
[manton] joined the channel
#
[tantek]
What's the bsky post limit?
#
[manton]
300. And it’s sort of baked into the definition of what a post is, so I have a feeling Bluesky will discard posts outside the limit.
paotsaq joined the channel
#
[manton]
I think there needs to be more work on having blog-like longer posts in AT Proto, in a way that is compatible with Bluesky ideally, not a completely new structure.
[dave] joined the channel
#
[tantek]
Huh, that's an odd limit. Like 20 more than the default Twitter 280? Weird
#
[tantek]
Mastodon's 500 limit sorta made sense as an easier to remember almost 2x of Twitter
#
[artlung]
Slack: 100000. Facebook 63206. LinkedIn: 3000. Myspace: 150. Discord: 2000. (If you are premium, 4000). X: 280 (If premium, 4000). GitHub PR comment: 65536.
#
[artlung]
(I made myself laugh by looking up the current MySpace limit). Also apparently Friendster is "back" https://friendster.com although. _Friendster Labs has no relation to or access to previous data hosted on http://friendster.com_
#
[tantek]
Hah those need to be added to the wiki [artlung] 😂
#
[artlung]
There are a few sites purporting to keep up with the limits. Though character counts in this context are very much "it depends" - how urls count, how a media upload counts against the limit, whether rich text works or not, how well multibyte characters work and whether they count against. All part of how the full stack processes stuff.
#
[artlung]
(one company I worked for had user messages with emojis getting eaten because of a database field that didn't allow it. the dba updated the mysql field to utf mb4 live but warned "there will be more to this" and indeed, it worked initially, but but multiple systems operated on those fields. backups and restores and replication and mimetypes and charsets and javascript libraries all needed tweaks and checks to "fix" it).
#
[schmarty]
"There will be more to this" is a good curse. Like "May you live in interesting times"
#
[tantek]
lol databases and encodings. Still can't get it right. Proves my point my storage vs caches (for dbs)
#
[tantek]
UTF8 flat text files FTW for longevity and data fidelity. Feels like that's a solved problem
#
[artlung]
My anecdote is from when people (users) started assuming that systems would have emoji support but many didn’t.
#
[artlung]
“ASCII ought to be enough for anybody“
#
[artlung]
most of the hello world CRUD stacks seem to have figured this out here in 2025
#
Zegnat
[artlung]: I always imagine that was why Mathias picked the pile of poo emoji for what I remember to be the main article about multibyte MySQL back in the days https://mathiasbynens.be/notes/mysql-utf8mb4
#
[artlung]
I think this is one of the articles that got shared when that company was dealing with this! And I remember reading it.
#
[artlung]
[Zegnat]++ nice pull!
#
Loqi
[Zegnat] has 4 karma in this channel over the last year (11 in all channels)
#
sknebel
yeah, apparently its easier to make developers care about emoji than about foreign languages ... at least sometimes
#
Zegnat
[artlung]: I remember having to open that one many times back in 2012
#
[tantek]
sknebel, perhaps corollary or consequence to easier to make developers care about what their users care (and complain) about
#
[tantek]
All these db/encoding problems would be relatively moot if these systems relegated databases to a caching layer, not persistent storage
#
[artlung]
Developers are not systems. My experience is the developers do care. That systems have limits and have inertia California, as polyglot and multicultural a political entity as I can think of, only in the last year allows, diacritical marks, accents, and tildes on official documents for names.
#
[artlung]
As developers, it is our responsibility to advocate for correctness and good data practices.
#
[tantek]
Sure, govs especially on intl interop, are largely still stuck in the 70s
#
[tantek]
[artlung] I was referring to the design and architecture of systems
#
[artlung]
Having worked in many terrible environments, I am sympathetic to developers who advocate for good kinds of change but are stymied by cost time, budget, etc.
#
[tantek]
There are two major problems (at least when devs are able to make decisions) : devs either (1) don't know how fragile databases are for storage, or (2) they deprioritize zero admin /longevity, or both
#
[tantek]
Right [artlung], either the devs are not empowered to make decisions, or they choose short term costs over longevity and long term costs
#
[tantek]
Sometimes choosing short term looks like cargo-culting
thegreekgeek and thegreekgeek_ joined the channel
#
immibis
if it's JSON-based, you can probably just add a new field for the expanded post
#
immibis
getting that standardized is more work, of course
#
immibis
All these db/encoding problems would be relatively moot if these systems relegated databases to a caching layer, not persistent storage <- then what is the persistent storage?
jonnybarnes and [snarfed] joined the channel
#
[snarfed]
filesystem
#
[snarfed]
or something equivalent like a cloud object store, S3 etc
gRegor joined the channel
#
[artlung]
My father just shared a photo of a quote from his kindle. It ends with “― Walt Whitman”
[aciccarello] joined the channel
#
[artlung]
The world is indeed terrible. And yet we must press on, and do our best to create robust systems that handle data well.
#
arekenaten
I'm not really bought into the idea of the database antipattern. I understand conceptually that a DB isn't the right choice for certain implementations, but unless your use case is already well covered (e.g. Static file serving) the DB adds a standard format for dynamics.
#
arekenaten
Is there a similar "templating language" antipattern for when you need to adapt content dynamically. Or am I missing something?
#
IWDiscord
<a​rekenaten>
#
[tantek]
there's nothing "standard format" about any database. W3C tried and gave up with Web SQL.
#
[tantek]
arekenaten, for templating there is such a broad spectrum of approaches that I don't know if any patterns have been well-identified, whether anti-* or good. There is a bunch of brainstorming also for improved approaches: https://indieweb.org/template
nemonical and nemonical1 joined the channel
#
immibis
S3 is a database. filesystems are databases. they're quite limited ones. why eschew the more capable database and force yourself to use the more limited one?
#
immibis
(performance is one reason, sometimes)
#
[tantek]
the "filesystems are databases" is a common misconception and trivializing of interop, as file systems have tons of interop, whereas databases do not and are typically constrained to their implementations.
#
[tantek]
it's basically a useless and misleading phrase
#
[tantek]
completely ignores user experience, owning your data, ease of migration etc. etc.
#
immibis
it's actually a common misconception that only SQL databases are databases databases
#
immibis
s/(databases) \1/\1/
#
immibis
things being databases has very little to do with "owning your data" and "ease of migration". database is to data as codebase is to code.
#
immibis
"database" is also often short for "database management system" or "database management software", such as mysql itself
#
immibis
postgresql or mariadb is database management software; software that is used to manage a base of data
#
[tantek]
completely missing the user significance and perspective and empirical experience with things that matter like longevity and fidelity over time. all that user-centric stuff that matters, not tossing around various dev-centric buzzwords
#
immibis
this is the dev channel, and both SQL databases and filesystems, when used as the backend of websites, are dev-centric
#
[tantek]
yep and none of that theoretical speak has any relevance when it comes to things actually working, for actual users, over a period of time
#
[tantek]
right, so it's ok to discuss them here, but as devs being user and UX first are part of our indieweb principles
#
[tantek]
we're not here to get CS PhDs
#
immibis
the user of your website sees neither the SQL database nor the filesystem
#
[tantek]
on the indieweb, you are the primary user of your website, and thus also the person responsible for longevity, migration, fidelity etc. all of which are harder and more fragile with databases from primary storage (as documented in the wiki) than flat file storage
#
immibis
no, i'm the primary developer. why would i need to make a website just to use it myself?
#
[tantek]
what is personal cloud
#
Loqi
A personal cloud is a software solution to administrate various cloud services, usually with a nice user interface, on servers you personally control https://indieweb.org/personal_cloud
#
[tantek]
^ lots of reasons why, lots of software for it, lots of people doing it
#
immibis
so what was your point about storage vs caches (for dbs)?
#
arekenaten
Also just to jump in and say because something "works" doesn't make it standard. There's all sorts of tricky interop problems that come with file formats between operating systems. Until you run into one it doesn't matter, but in the niche case you do it's annoying to have to translate. Similarly to DB nuance between SQL implementations in my mind.
#
arekenaten
Line endings alone are enough to merit workarounds that I know of off the top of my head so I'm sure there are more.
#
Zegnat
I guess the interop thing is "a continuous line of bytes supposed to be parsed according to the unicode standard" ... and then it will still turn out that not everything can render all the code points to something visible. But that at least is getting better every day. (I guess then only endianess is still going to be a bother? Not sure the consortium does anything with that.)
#
arekenaten
Regardless, I feel like there's a whole lot of wiggle room in the "should I store everything as flat files" question. Antipattern is a strong word imo.
anon1 and gRegor joined the channel
#
pfych
Amazon runs entirely off a json file stored in s3 technically 😉 (dynamoDB)
#
[tantek]
anyone remember which IndieWebCamp sessions we discussed doing local development setups? like many years ago, prepandemic
barnaby, ttybitnik, [timothy_chambe], alephalpha0, gRegor and gRegorLove_ joined the channel