#dev 2022-02-15

2022-02-15 UTC
jacky and lagash joined the channel
#
[tantek]
wow lots of scrollback
#
[tantek]
good discussion about exporting / exportability
#
[tantek]
interesting that the "RSS export" code path in Known sounds like it rotted (because who was checking it or viewing it? how often?) in contrast to the visible HTML+mf2
#
[tantek]
also why is "flagship implementation" even a good question to ask / framing to assume?
#
[tantek]
to me that's kinda like asking is there a monoculture implementation?
lagash, dovedozen[d] and [tonz] joined the channel
#
sknebel
capjamesg[d]: what level of detail do you want? ๐Ÿคฃ
[jeremycherfas] joined the channel
#
capjamesg[d]
Murray++ for the guidance! Regarding links, I like the idea of having them so I can link to a particular point on a page.
#
Loqi
Murray has 4 karma in this channel over the last year (6 in all channels)
#
capjamesg[d]
But I suppose a table of contents in a details/summary section with those links would suffice.
#
capjamesg[d]
But then there's accessibility to consider. On a page with six headings, do I want all of them to be read out by a screen reader before the reader actually gets to the main content in the page.
#
capjamesg[d]
sknebel I shall review the zine linked above and maybe come back ๐Ÿ™‚
jjuran joined the channel
#
capjamesg[d]
Murray that guide was really helpful. Thank you!
#
sknebel
capjamesg[d]: ok :D I was nerd-sniped into opening the kernel sources before I stopped myself and asked first, so ... ;)
#
capjamesg[d]
I think I have a better understanding now. But I still don't understand the actual role the OS plays. Is it right that there is basically a hash map between services and all valid internal IPs and ports, one for UDP and one for TCP? Or something along those lines? When an application connects to port, say, 80, all those connections are routed to the corresponding service?
#
capjamesg[d]
This starts to explain how NGINX proxy passes work to me which is quite exciting.
#
capjamesg[d]
To the level I understand, NGINX routes my HTTP request through to the application socket (say, microsub.sock) where my application is listening and will process the request.
#
sknebel
not just basically a hash map, literally a hash map :)
#
sknebel
of course additional earlier stages that deal with "this is a packet on an ethernet interface. is this packet even for this computer?" "ok it is, is it ipv4, ipv6, ...?" "is that an IP I use?" "is it TCP/UDP/...?", but then "ok, which sockets are associated with that port"
#
sknebel
(ok, that's a detail: its not explicitly mapped to an app, but to a socket - which is owned by an app)
#
sknebel
(socket API is worth a look too as low level "how do I talk to the OS about this stuff")
#
capjamesg[d]
IndieWeb Search returned three relevant results for this query ๐Ÿ™‚ https://indieweb-search.jamesg.blog/results?query=unix+sockets
#
capjamesg[d]
That makes me so happy!
#
sknebel
heh. unix domain sockets are actually a special case that doesn't quite apply
#
sknebel
also useful to know though
#
Murray[d]
capjamesg[d] glad it was useful ๐Ÿ™‚ In terms of concerns over tables of contents, as long as they're given a clear heading and are marked up as a list, anyone using any form of assistive tech (AT) will have browsing habits that can skip past them if not wanted. E.g. lots of screenreader users browse by heading or section, so as long as a contents table is distinctly labelled, they can just skip clean over it. In many ways, providing a skele
#
Murray[d]
skeletong/skeleton ๐Ÿ˜‚
#
capjamesg[d]
sknebel yes it was part of a separate itch I had to scratch in my mind ๐Ÿ™‚
#
capjamesg[d]
If anyone has any feedback on IndieWeb Search I'd love to hear it. I think I have worried too much about relevance engineering given what data one can access as a niche search engine operator.
#
capjamesg[d]
Murray that's good to know!
#
Loqi
Murray has 5 karma in this channel over the last year (7 in all channels)
#
capjamesg[d]
sknebel++
#
Loqi
sknebel has 9 karma in this channel over the last year (36 in all channels)
king joined the channel
#
king
hello
#
king
With our IRC ad service
#
king
rampjoul
#
king
free money
tetov-irc, jamietanna, jacky, KartikPrabhu, [KevinMarks] and Eddy04[d] joined the channel
#
capjamesg[d]
sknebel What do DNS authoritative name servers like bind implement?
#
capjamesg[d]
That is absolutely crucial for the average DNS query that is.
#
capjamesg[d]
If I understand correctly, a DNS server only needs to return a response on port 53 to a query like "find me all of the A records for jamesg.blog please" (obviously not in that language).
#
capjamesg[d]
That response is structured in a way that the server looking up the name server can understand.
#
capjamesg[d]
The reason I ask is that I'd love to implement one with Python's dnslib library. It handles the responses but I need to build the logic ๐Ÿ™‚
#
sknebel
pretty much, yes. probably requests for A, AAAA and ANY, thats I think pretty much it to just serve a HTTP domain
#
capjamesg[d]
Oh any I hadn't thought about that.
#
capjamesg[d]
Can I send over some code to glance at?
#
capjamesg[d]
To trigger, you can use a command like: dig @127.0.0.1 -p 5005 jamesg.blog TXT
#
capjamesg[d]
ANY queries don't work yet haha.
#
sknebel
I'd try with a subdomain for testing first before I risk my emails :D
#
capjamesg[d]
Good point.
#
capjamesg[d]
What should you do if the server can't respond to a query?
#
capjamesg[d]
Like if you try to query google.com using my server?
#
Loqi
A toy DNS resolver
#
sknebel
that's exactly *not* what capjamesg[d] wants :D
#
sknebel
capjamesg[d]: https://powerdns.org/hello-dns/auth.md.html is a good starting point
#
sknebel
(I knew berthubert had written about that stuff, just took some time to dig out)
barnaby joined the channel
#
capjamesg[d]
[KevinMarks] I have seen that but it looks like a task for another day.
#
capjamesg[d]
Thank you sknebel! I did try and read the spec a few weeks ago but it was a bit verbose for me at my current level of knowledge and understanding of networking.
#
sknebel
DNS specs also are ... not fun
#
sknebel
spread over a lot of documents and iterations
#
sknebel
but the first few paragraphs here have already details I wouldnt have remembered top of my head :D
#
sknebel
("a lot": https://powerdns.org/dns-camel/ - "There are 297 RFCs relevant to DNS, of which 108 are selected by filter. Total pages selected: 2082")
#
sknebel
(although a large pile is probably stuff like DNSSEC which isnt actually that relevant for most)
#
capjamesg[d]
jvns.ca has implemented her own DNS server but only for a side project.
#
Loqi
New tool: Mess with DNS!
#
jacky
I feel like I'd 100% want to play with DNS to resolve people's p-nickname to `.contacts`
#
jacky
first my /contacts, then falling back to /chat-names lol
#
capjamesg[d]
Running my own DNS server does mean I can send arbitrary TXT records programatically. Not that I'd really want to do that for anything important but it's a cool perk.
#
capjamesg[d]
For instance my DNS server could rotate a TXT record with a different type of coffee each day. Again, not important stuff.
#
aaronpk
I have a dns record that reports what city I am in
#
aaronpk
but really it's just a cron job that updates the record at the Linode DNS API so not as cool as actually dynamic
#
jacky
capjamesg[d]: if it's important to you, that's all that matters! :)
#
jacky
didn't realize how handy/capable DNS is tbh
#
[KevinMarks]
dns-sd is pretty neat
#
barnaby
aaronpk: what kind of record is the location, and on which domain? I donโ€™t see it in `dig aaronparecki.com txt`
#
aaronpk
dig txt where.is.aaron.pk
#
barnaby
ha ha very nice
#
capjamesg[d]
My reservation right now is security.
#
capjamesg[d]
And also whether what I have is roughly stable.
#
capjamesg[d]
sknebel Apologies for all the questions but are there any big security issues that one should be aware of building a name server?
#
capjamesg[d]
I'm deferring the actual header production to a library so I'm not worried about payload formats.
#
sknebel
the hello-dns link had a good point about preventing amplification attacks by checking queries are queries etc. there might be more in that direction
#
sknebel
otherwise I'd guess something in python with little complex backend logic is reasonably safe against the usual dangers
#
barnaby
btw capjamesg[d] if youโ€™re looking to learn more about networking, teach yourself CS recommends a few good resources https://teachyourselfcs.com/#networking
#
jacky
question for people's Micropub implementation (and a bit of a question on Micropub clients): do y'all do `u-*` property resolution of incoming things? I figure for servers, one _could_ do this if URLs are provided but I wonder if this is something clients _should_ do to make the act of publishing easier
#
jacky
easier in the sense that a server wouldn't have to work to normalize external URLs into inline MF2 (if they choose to do so)
#
sknebel
I dont get what you mean
#
jacky
so like let's say I use a client and it sends just h=entry&like-of=https://indieweb.org/why
#
jacky
currently, in my implementation, it takes that as face value, lightly munges it into MF2-JSON for storage and then when I render that page, I use those properties to build the page back up
#
jacky
like that string would be the conventional `{"type":"h-entry", "properties": {"like-of": ["https://indieweb.org/why"]}}`
#
jacky
what I'm asking if the act of turning `https://indieweb.org/why` into something like a h-cite should be done on the client or the server side
#
aaronpk
i do that on the server side
#
aaronpk
expanding like-of and in-reply-to
#
barnaby
IMO server side makes the most sense, to keep publishing UI and code simpler (i.e. a text box for a URL rather than an interactive h-cite builder)
#
sknebel
usually done server-side, but I remember some discussions about doing it client-side, so there might be a client somewhere that does it
#
barnaby
so a micropub source query would just return the original like-of URL, but on your site you can mark it up however you want
#
jacky
Server-side seems like the consensus
#
aaronpk
Quill does preview reply URLs if it can, but it doens't send anything other than the plain URL
#
jacky
I guess my _other_ hope is to give things like static sites a richer reply context (which is also where I think it's a bit of a win-win client-side impl wise - I don't have to add extra logic for specific properties, the server can be a bit more 'dumb')
#
sknebel
not sure how thats related to static sites
#
jacky
the richer reply context bit
#
aaronpk
The micropub server for the static site can still do that work
#
sknebel
but the question client/server doesnt care if the rendering later is static or not?
#
sknebel
(barnaby: source query is interesting, because one could offer the abilitiy to edit the preview. but agreed, havent seen that)
#
jacky
aaronpk: true
#
jacky
I guess it's really me needing to pick where I want to put more of the 'work' to do that
#
barnaby
agreed, there might be valid use-cases for editing h-cites in micropub clients. I wonder if itโ€™d make sense to establish a convention (if it doesnโ€™t already exist) for allowing clients to either handle full embedded mf editing if they can, or falling back to a simple string
#
barnaby
maybe something like if a source query returns a property with an embedded h-cite on a like-of, a client could either offer an h-cite UI, or look in the 'value' key under the h-cite and take that as the plain text value
#
jacky
barnaby: is that in the case of editing or creating? just making sure I follow
#
jacky
I can see that for editing
#
barnaby
as the value key is already established in mf2 as a plaintext fallback
#
[tantek]
One key use case is the ability to edit the reply context on the client side instead of what we reply context is created / expanded by the server, eg to trim or otherwise edit what is shown (hiding spoilers, or CW related materials etc)
#
jacky
but in the case of creating, would it be sending the value of `u-like-of` (or what have you) _to the Micropub server_?
#
jacky
[tantek]: that's a good example
#
barnaby
jacky: Iโ€™m specifically talking about editing in this case, as Iโ€™m assuming that if a rich h-cite creation UI exists, then that client is also going to support editing it
#
[tantek]
Same thing with editing link previews in general for /like or /bookmark posts. You might not want the image or specific summary paragraph provided by the author of the destination link to show up on your post
#
jacky
barnaby: that makes sense
#
jacky
[tantek]: yeah (though I can see that being used for rickrolling people, lol)
#
[tantek]
Some link preview images are either ugly (some CW) or duplicate the icon of the author (thus it looks like a bug on your site when you display the โ€œsameโ€ image twice), and some link preview summary text is site spam stuck into OGP metacrap โ€œthis is the best site for photos!โ€ Etc
#
jacky
I think I might just need to either work on a Micropub client or contribute to one to experiment on a living example
#
barnaby
and if thatโ€™s the case, it might be worth establishing a convention to let clients edit embedded mf structures as their plaintext value without having to implement specific UIs for absolutely everything, perhaps with a warning, that some data might get lost in the process
#
jacky
barnaby++ on that bit
#
Loqi
barnaby has 8 karma in this channel over the last year (15 in all channels)
#
[tantek]
A micropub client should show you a link preview of any links and then allow you to edit those, and if you do then sends the edited version to the server as an h-cite, which the server should respect instead of doing its own link preview extraction & display logic
#
jacky
one takeaway I _am_ getting from this though is that it's never a bad idea to handle this processing on the server
#
jacky
(which was my initial question but I got very curious about the client experience as well)
#
[tantek]
The answer to which, is both client & server. The server should handle / do reply contexts and link previews by default, and the server should respect when a client provides a UI to edit those and sends across a full h-cite instead of just the link
#
jacky
noted!
#
barnaby
that would either require every micropub implementation to support h-cites, or a way of querying what properties the server supports to avoid undermining users expectations (say, when the client sends a like-of h-cite rather than string, and the server either doesnโ€™t support it or makes its own link preview)
#
jacky
I think there's an extension to report the structure of properties that a Micropub server supports
#
jacky
I think jamietanna has a notable implementation
#
barnaby
which reminds me of the idea of building post previewing into micropub, so that users can immediately see which edits their server actually responds to, and can decide whether to go through with posting something based on how *the actual server theyโ€™re posting to* is going to display it
#
barnaby
oh that sounds cool, Iโ€™ll check it out
#
jacky
ah yeah that is
#
Loqi
[jamietanna] #33 Query for supported properties, for a supported post-type
#
Loqi
[manton] I've implemented a version of this in Micro.blog to see how it would look. It returns the supported properties directly from `q=config` without needing any extra requests. My initial implementation looks like this, although it's probably missing some...
#
jacky
hmm but that's mainly for post-types and not like container items (like h-entry, h-cite, etc)
#
barnaby
hmm that covers supported properties, but nothing about nested structures
#
barnaby
but it might be easy enough to add that in
#
jacky
going to muse on this over the week, barnaby, your point got me wondering if this is something to report or if there's some way to work around
#
barnaby
IMO establishing the convention of MP clients and servers always providing the value key on nested mf structures as a plaintext fallback would mostly fix this
#
jacky
true that's the easiest way too :)
#
barnaby
more detailed capability querying could definitely be useful, but having an established fallback convention is useful even in that case
#
jacky
guessing this is something to document on the Wiki on /Micropub
#
jacky
I'll try to capture this idea there
#
barnaby
and allows for quicker development of both clients and servers, as people can start off by just implementing the plaintext version, knowing that they can add a more advanced implementation later without breaking anything
#
barnaby
sounds good jacky++
#
Loqi
jacky has 20 karma in this channel over the last year (62 in all channels)
[manton] joined the channel
#
[manton]
I wonder if it would help to document which clients actually pay attention to the โ€œpost-typesโ€ โ†’ โ€œpropertiesโ€ field (as documented in that GitHub issue). Off the top of my head, Iโ€™m not sure I know of any, but it has been a couple years since I added this on the server.
#
barnaby
mf data handling already established the de-facto principle of โ€œif you expect a property value to be a string but you get something more complicated, use its `value` keyโ€ for nested mf, embedded HTML and most recently img alt parsing
#
jacky
[manton]: good point
#
jacky
might need a micropub client bash where some people just test them all out to see where they are and how they match up with the specs (could take a while, lol)
#
[manton]
Heh, yeah.
#
[manton]
I also think thereโ€™s a bit of a disconnect between clients that will think about MF2 at all, vs. most that will just send โ€œnameโ€ and โ€œcontentโ€ and call it a day.
#
jacky
ha - the MVP of clients (sometimes that's all that's needed)
#
barnaby
IMO thatโ€™s one of the great things about micropub! the barrier to entry is so low
#
barnaby
so itโ€™d be great to find ways of continuing to allow clients like that to function while allowing more complex clients to add additional features
#
[manton]
Yeah. And my 2 cents, I worry about complicating Micropub with complicated structured posts, since it risks scaring off basic implementations.
#
[tantek]
barnaby, the fallback is already in the mf2 property value even when it has an embedded object
#
[tantek]
This is how a <span class=โ€œu-like-of h-citeโ€>โ€ฆ<a class=โ€œu-urlโ€ href=โ€œโ€ฆโ€> works
#
barnaby
[tantek]: yeah thatโ€™s what I mean! and hence IMO it makes sense to encourage micropub clients and servers to publish and consume the value property (which afaik is not currently mentioned anywhere?) so that people can continue to use this familiar, useful pattern
#
[tantek]
The u-like-of gets its โ€œplain propertyโ€ (not just plain โ€œtextโ€, it depend on the prefix of the property) of the URL from the u-url of the embedded h-cite
#
barnaby
e.g. in the nested microformat object section of the spec, it doesnโ€™t show the value key in the example https://www.w3.org/TR/micropub/#nested-microformats-objects
#
barnaby
[tantek]: yes I know, and itโ€™s really useful! my suggestion is that this pattern could be useful in micropub too, but afaik isnโ€™t currently being used or encouraged
#
[tantek]
That should be updated in micropub.spec.indieweb.org then to be consistent with mf2 parsing
[snarfed] joined the channel
#
[snarfed]
this all reminds me a bit of https://snarfed.org/2015-11-29_keep-bridgy-publish-dumb , in which people gradually asked for more and more rich interactivity from Bridgy Publish, which I largely refused, and which Micropub clients and servers are maybe more suited for
#
[snarfed]
(different use cases, but related)
#
aaronpk
that particular example uses "summary" for that instead
#
barnaby
I need to get back to work but later on I can write up some MP issues about this
#
barnaby
aaronpk: which would be fine if publishing and looking for the โ€œsummaryโ€ property was established as a fallback pattern, but it isnโ€™t, and doesnโ€™t work in the case of embedded like-of/reply-to h-cite we were talking about above
#
barnaby
whereas looking for a value key is already established, and is quite generalisable to all embedded mf structures
#
barnaby
also the summary is a property on the top level structure, not a way of editing the nested h-measure structures, which IMO are the ones which should have value keys as fallbacks for editing. although admittedly this example is specific enough that users are only likely to want to edit it using a use-case-specific app
#
aaronpk
yeah i would want to discuss this in the context of some simpler structures
#
barnaby
but even in this case, with value keys a simpler client could see that there are these weight and bodyfat properties which it doesnโ€™t understand, but which have value keys (which could be, say "70.64kg", and could offer the user the oppotunity to edit them as plain strings, with a warning that data might get lost. and then the server, on receiving the data, could either ignore it as itโ€™s not the expected nested structure, or
#
barnaby
try to parse it into one by looking for a number and a unit
#
jacky
hmmm this is def deeper (and interesting) dialogue
#
Loqi
[cillian64] GitHub leaked names of private repos with pages
#
capjamesg[d]
I was quite concerned when I first saw the title of the email which made me think that I had exposed some kind of private information by accident.
doosboox8 joined the channel
#
jacky
that also introduced me to https://phish.ly/
#
[schmarty]
missed the "rich h-cite editing" discussion, but i actually prototyped adding this to Quill at some IndieWebCamp https://github.com/aaronpk/Quill/compare/main...martymcguire:mm-nested-h-cite
#
[schmarty]
it was some q&d contenteditable stuff but it gave me some hope.
#
[schmarty]
my opinion on client vs server unrolling of urls-to-reply-contexts is: offer it all because it comes down to preference and circumstance.
#
[schmarty]
(and reminds me that there are a handful of mini micropub clients i'd like to see that are purely for editing bits of posts. editing tags, unfurling/editing reply-contexts, updating syndication, adding media, ...)
edburns[d] joined the channel
#
jacky
this looks like an interesting editor (for Linux-y systems) https://thiefmd.com/
#
jacky
meant to provide a Ulyssess-style editor
#
jacky
got curious because it has support for Wordpress
[tw2113_Slack_] joined the channel
#
capjamesg[d]
aaronpk You posted a story to your site! I have been waiting to see that function in action for a while.
#
aaronpk
oh crap i did haha
#
aaronpk
i was trying to post a video
#
capjamesg[d]
It was an accident? ๐Ÿ˜‚
#
capjamesg[d]
From #indieweb. I think I need to start taking into account internal links in ranking.
#
capjamesg[d]
But I don't currently calculate that in the link graph calculation.
#
aaronpk
i forgot that it would actually post the story, all the video uploads i was trying were failing
#
barnaby
maybe sniffing a common prefix/suffix for a particular domain ("Barnaby Walters โ€ข" in my case) and strongly down-weighting it could help improve things, if thatโ€™s possible?
#
barnaby
afaik itโ€™s a very common pattern in <title>s
#
[schmarty]
i think i don't understand how the indieweb search works. only seven results for caturday and none of them from https://martymcgui.re/tag/caturday/ ?? ๐Ÿ˜‚
#
barnaby
turns out making a good search engine is a hard problem :D
angelo joined the channel
#
[schmarty]
(fwiw i include the word "Caturday" in the content for each post, along with the tag "#caturday")
asteria joined the channel
#
jacky
might be interested for people with audio snippets on their site and wanna visualize it somehow https://css-tricks.com/making-an-audio-waveform-visualizer-with-vanilla-javascript/
asteria, Seirdy, angelo and capjamesg1 joined the channel
#
capjamesg1
Well Discord is down ๐Ÿ˜ฑ
#
capjamesg1
barnaby all the code is open source if you want to assist in any way :)
#
capjamesg1
It is indeed a very hard problem. After embarking on this project I realized just how much google needs to do behind the scenes.
#
[schmarty]
jacky: oo i like that it's very how-to and the result is aesthetic. kinda want to render them to an SVG instead
#
capjamesg1
[schmarty] crawls were initially capped to a small number of URLs so your site might have been affected by that. Iโ€™ll run a full recrawl of your site tomorrow.
#
[schmarty]
(an image proxy that eats mp3s and spits out pretty waveform SVGs ๐Ÿค” )
#
capjamesg1
The common prefix thing is a big issue. Google does rewrite title tags interestingly.
#
[schmarty]
capjamesg1: ah, good to know! is there a way to see what the search engine "knows" about a given site?
#
capjamesg1
Ask me :) there will be a JSON API that lets you see incoming links to a page, internal links, and all the data the engine has in each page on each result (minus a few things)
#
capjamesg1
I just need to push a fix because it was affected by the refactor
#
capjamesg1
What do you want to know? I would love to make a site owner dashboard so you can see stats but do Iโ€™d love input on that.
#
capjamesg[d]
Discord is back!
#
[schmarty]
capjamesg: the first and only thing so far is number of pages crawled and, like, latest-page-crawled time.
#
capjamesg[d]
That is doable. Any other feedback on the engine? Iโ€™m actively chipping away at it and am happy to introduce / change ranking factors etc.
#
[schmarty]
capjamesg++ i've mostly been watching the project from afar and haven't tinkered much, yet. very cool stuff!!
#
Loqi
capjamesg has 29 karma in this channel over the last year (75 in all channels)
#
jacky
[schmarty]: tbh same! wouldn't be surprised if there's a library for that
#
capjamesg[d]
Bad results pages help me figure out how to improve things. There are so many moving pieces that I havenโ€™t done as much with relevance as I know I should.
e-snail joined the channel
#
[schmarty]
that helps make it feel like glitch is a too-magical irreplaceable thing to put projects on.
#
[schmarty]
err sorry. makes it feel like it's _not_ too-magical lol.
#
kinduff
haven't tried DigitalOceanโ€™s App Platform, but sometimes i feel it's kinda costly
[metbril] joined the channel
#
kinduff
5 dollars a month is nice but a droplet for 10 dollars has more transfer and no cpu/build minutes
#
kinduff
but happy to see the glitch community deploying their apps
jacky, [metbril], [James_Van_Dyne], [snarfed], [manton], [tw2113_Slack_] and [jeremycherfas] joined the channel
#
jacky
[schmarty] took months to get this out (still consulting with them)
#
[schmarty]
jacky++ i can imagine!
#
Loqi
jacky has 21 karma in this channel over the last year (63 in all channels)
#
[schmarty]
whew okay trying to resolve dependabot issues in my indieweb projects on glitch is giving me the sads. kapowski is presently broken and thankfully i decided to touch it before the webring. ๐Ÿ˜
#
Loqi
npm has -1 karma in this channel over the last year (-2 in all channels)
angelo joined the channel
#
barnaby
okay I typed up the micropub value key fallback suggestion from before, with some fully worked-out examples https://github.com/w3c/Micropub/issues/124
#
Loqi
[barnabywalters] #124 Consider defining use of `value` key in nested microformats for fallback
#
aaronpk
just saw that
#
aaronpk
not sure how i feel about that, but good for them
#
@QuinnyPig
Is Linode going to modernize Akamai's API, or is the Linode Terraform provider going to be refactored to use Mechanical Turk to open a ticket with Akamai's existing "Jason API?" https://twitter.com/jordannovet/status/1493694280375967745
(twitter.com/_/status/1493709992515432450)
sebbu, rhiaro, barnaby and tetov-irc joined the channel
#
[tantek]
[schmarty]++ yes that was at IWS 2019! And IIRC after a long side-conversation you and I had during create day on that very subject and related: how to store information from or about other people's sites (e.g. from webmentions, or reply-contexts / link-previews, which are different because of who external vs author initiated), how to allow the site author to edit them, perhaps before posting (or moderating a response), and how to
#
[tantek]
display them
#
[tantek]
I'm still stuck on those storage questions of "not my data" tbh
#
[tantek]
my suspicion is that I want to keep separate storage / caching policies (and files perhaps?) for different levels/kinds of trust/authority/responsibility
#
[tantek]
at the top (1) is content I've written myself, then perhaps (2) reviewed/edited content extracted from the URLs that I've written (e.g. edited/reviewed reply-contexts/link-previews), then (3) raw content from the URLs that I've written (automatic reply-contexts/link-previews),
#
[schmarty]
Those feel like good dividing lines to start from!
#
[tantek]
then Webmention received stuff, i.e. (4) edited/reviewed content from the URLs from folks in my nicknames cache (sort of applying /Vouch principles to content storage), then (5) edited/reviewed content from other URLs (not in nicknames cache), then (6) raw content from URLs from nicknames cache, then (7) raw content from other URLs
#
[tantek]
one point being, the further down in those levels, the more I'm ok with "losing" the "content" and re-retrieving it (treating it like a cache)