#dev 2020-04-25
2020-04-25 UTC
[tantek] joined the channel
# 
KartikPrabhu that would make mf2 in some page dependent on having an entire another page for each h-*
# KartikPrabhu seems unreasonable. Specially with static pages, it would require having the synchronize the data between the page and the URL of the h-* object
# KartikPrabhu or put an id on each h-* object
# KartikPrabhu I don't see why we should put such burden on publishers
# 
jacky like if my h-card only lives at https://jacky.wtf/#hCard, that doesn't seem any more complicated
# KartikPrabhu it *can* be done. But requiring is a bit much
# KartikPrabhu sure. until you find a case where that strategy doesn't work, that is fine to use
# KartikPrabhu bullet-proofing against something that just might not happen is not useful IMO
[TimApple] joined the channel
# 
[TimApple] Hello, I've been lurking a bit. Was looking for recommendations on web hosts.. they all seem the same to me. Right now my plan is to do everything myself using vanilla javascript. Any suggestions on hosting provider would be great. Thanks.
# [TimApple] Sorry, think i should of put this in the indieweb channel.
# 
Loqi Web hosting can be the primary regular cost in maintaining an IndieWeb site; this page lists several options from free on up depending on your publishing needs, like a static, shared, private, or dedicated server https://indieweb.org/webhosts
Aishawhite, [tantek], dckc and [LewisCowles] joined the channel
# [LewisCowles] jacky, when publishing information from non internet systems in read-only format (what I'm doing now); the id and a single URL just are not important
# [LewisCowles] I know what you mean, and in an ideal world, I'd publish the private systems id and url because it wouldn't need to be private, but I think a worse case scenario is publishing details of a private setup others can't even reach
# jacky I think in your case mentioned here, https://chat.indieweb.org/dev/2020-04-24#t1587762799673600, it's fine to not even have HTML for the page
# Loqi [[LewisCowles]] if a u-url is not public, or cannot be known at time of generation. Should it be omitted?
# [LewisCowles] Right, but I want to have the HTML, I just am not sure of the value of providing a u-url or u-id people cannot interact with past consumption
# [LewisCowles] h-feed makes sense, as will h-entry and h-card
# [LewisCowles] similarly I'm not u-uid'ing or u-url'ing on any of those sub-entries.
# [LewisCowles] In order to do that in a way I'd feel comfortable with I'd need to A, Identify members of a closed system elsewhere (small problem right now as it's just me)
# [LewisCowles] B, provide the canonical u-uid and/or u-url for them, and the resources on their source
# [LewisCowles] otherwise I'd be depriving those users of their independence
# [LewisCowles] or needing to facilitate some form of mailbox for others / other systems resources on my own
# [LewisCowles] distributed systems: Share the least surface area practically possible between two systems.
# [LewisCowles] it's both RE: sub-net and authorization. It's unlikely without hacking into one of my multiple home networks that you'd be able to get in, and then you'd need to pass about 4 layers of security
# [LewisCowles] having my public IP also wouldn't help as I don't externalise any of the systems
# [LewisCowles] it's about selectively, being able to output that content to the web
# [LewisCowles] in static form
# [LewisCowles] then transform that for web use-cases
# [LewisCowles] such as subscription to a kanban board to monitor progress until it's complete
# [LewisCowles] I suppose I could make the u-url the one on my site, but that was the point from last night at which I began to sigh. I'll need to build it as part of deployment. Only at deploy time should I know that URL and it's not intrinsic to exporting a KanBan board from the system I export from
# [LewisCowles] So at deploy time, parsing a HTML document (jsdom looks most promising) and transforming it, uploading it's transformed output for a number of pages)
# [LewisCowles] it's just a lot to unpack
# [LewisCowles] KevinMarks, I just saw your comment about relative urls. I suppose I could use one of those for the feed
# [LewisCowles] since I do control and mandate the file-system layout at point of export, just to keep the CSS, HTML and assets hanging together that might be a great way to package this.
# [LewisCowles] tantek++ RE: placating validators
# [LewisCowles] as an example, testimonials were once marked up as reviews for a customer. Well reviews have a rating, so what did every review get (although I gave the customer control to set their own rating). Everything is 5-star
# [LewisCowles] And these were from regional companies, so large-ish names for a regional-provider company to showcase to boost authority and get some internet juice.
# [LewisCowles] yeah tbf I got an alert yesterday that something from 2014 no longer matched google's requirements for a job advertisement. They now require salary, location some other data which makes sense, but that they didn't need
[KevinMarks] joined the channel
# 
[KevinMarks] Why can't you use a relative u-url?
# [KevinMarks] That fits the bundle and dat cases
# [LewisCowles] ???
# [LewisCowles] > KevinMarks, I just saw your comment about relative urls. I suppose I could use one of those for the feed
# [LewisCowles] I'm totally entertaining the possibility of exporting with a relative u-url
# [KevinMarks] Also has the advantage of working locally with file urls
# [KevinMarks] It is a trade off - one of the problems with atom feeds etc is relative urls. Adding a base at serve time can help. Otherwise what is fetching may need to preserve headers.
# [LewisCowles] I suppose a relative URL could assist a deploy-time way to enhance properties to use non-relative URL's
# [LewisCowles] it's a very involved process
# [LewisCowles] I may well come out the other side with "here's why I decided to side-step" parts or the whole thing
vika_nezrimaya joined the channel
# [LewisCowles] I want to be able to export database and app-reliant things to static output. I have that part done, submitted to upstream for approval.
# [LewisCowles] Enhancing it is a nice side-step that could enhance the ability to get the data back out, which is also part of my initial to-static in the form of a proprietary JSON format
# [LewisCowles] the JSON is in-fact easier to export en-masse with a list of board PK's, but it's inert
# [LewisCowles] the HTML I feel represents something of-use in a broader range of contexts. Which micro-formats could help with.
# 
Zegnat !tell jacky from yesterday’s call, I think this is what both PHP and Python parsers have tested with as a possible whitespace cleanup alg: https://wiki.zegnat.net/media/textparsing.html
# [LewisCowles] Zegnat is this just to make pretty JSON?
# [LewisCowles] because HTML ignores whitespace. I was blinking a fair bit when modifying whitespace came up
# [LewisCowles] so you're only running it on p-* types, not e-* types
# Zegnat We kind of want to approximate what a browser would give you when calling .innerText (https://html.spec.whatwg.org/multipage/dom.html#the-innertext-idl-attribute)
# Zegnat But turns out implementing HTML innerText (different from DOM textContent!) is a bit of a slag when you aren’t a render engine (ie. browser): https://github.com/Zegnat/php-innertext
# Zegnat microformats2 spec discussion https://github.com/microformats/microformats2-parsing/issues/15
# Zegnat implementation details in PHP parser https://github.com/microformats/php-mf2/pull/168
swentel and Florella joined the channel
# [LewisCowles] I've unfortunately necro'd on that GitHub issue
# [LewisCowles] it seems to be a case where following a spec is problematic because it's too specific, but when I threw in some HTML from a source. I didn't get the output I'd expected
# [LewisCowles] the spec marking as a computed field, even if cached computed from the saved HTML would make a tonne of sense to me
# [LewisCowles] this is a projection of whatever algorithm fits your use case to text-ify (pardon that marketing jargon BS) rich content as less-rich content
# [LewisCowles] I've also hinted that plain-text might have a variant depending on application
# [LewisCowles] so also it would mean in taking "text" from a h-entry e-content or similar on your site. I'd know, I either had to go back to the source (your site) at some arbitrary period (as algorithm improves); or rely on a pre-cached part as e-content already provides
# [LewisCowles] jacky would then have a very good case for mandating u-url props for that case
# Zegnat The idea with mf2 is to get structured data ready to work with in a way that abstracts away the HTML parsing. So people do not want to go back to having to parse the HTML themselves to find innerText. Makes sense to then make the argument for switching from textContent to innerText in the mf2 spec (which is that issue)]
# Zegnat Yeah. I think this particular case comes down to people expecting innerText and getting textContent though. innerText is what is expected because it is what browsers would display, even text browsers like Lynx. textContent is much more like serialised AST output and therefor part of the DOM standard.
joshghent joined the channel
# [LewisCowles] lynx in this case displays very differently
# [LewisCowles] I'll attach
# [LewisCowles] basically I'm not knocking the algorithm or the work
# [LewisCowles] I'm suggesting that there may be an as-yet unconsidered pivot in spec
# [LewisCowles] when or how-often text is regenerated is mildly unimportant
# Zegnat I wonder how https://github.com/Zegnat/php-innertext would hold up against that piece of test HTML you ran there, [LewisCowles]. As that actually tries to implement innerText
# [LewisCowles] I may run it for fun
# [LewisCowles] without writing that it might be that `text` is not very useful at all, and should be dropped
# [LewisCowles] Many years ago (2009-2010) I had a project from a business my sister was group operations director for. They were sending image-only emails wondering why response rates were < 1%
# [LewisCowles] To me it's natural to think. "Oh if people can't read it, then it won't matter how well targeted your communication is"
# Zegnat Thing is that the issue originated with heavy mf2 consumers, no with the spec creators, so clearly there is a need for this in application context. And if we can address a lot of those needs by running minor white space cleanups, I feel like we should try and spec those necessary cleanups for all parsers
# [LewisCowles] Did a bit of work on creating alt-text fallback for those, and HTML content for those with text part fallback
# [LewisCowles] in the end > 80% of people read the emails. Because the issue in creative presentation had been addressed
# [LewisCowles] you're right in that it makes parsing more expensive. I had the benefit of a manual process
# [LewisCowles] without valuable text content, we're just pushing bytes to people
rhiaro_, rhiaro, [TimApple] and [jgmac1106] joined the channel
# Zegnat GWG my default PHP testing and linting setup right now, would love to hear if you have any improvements for me! https://github.com/Zegnat/php-website-starter/blob/5adf61a211b927d1b1d8de6bcefae15696892aa3/composer.json#L24-L33
# [LewisCowles] 😢 https://github.com/PuerkitoBio/goquery source documents must be UTF-8.
# [LewisCowles] I can guarantee this for me. I can wish it for others. I can't force it and may have difficulty spending time and effort ensuring it.
desmondrivet, globbot, MylesBraithwaite, mlncn, nloadholtes, [tantek], aaronpk_, thebaer__, Verlf, chrisaldrich, twMat, brunow, doubleloop1, [prtksxna], [chrisaldrich] and BillSeitz joined the channel
# Zegnat aaronpk: does your picrofeed fork have a replace option in its JSON? https://getcomposer.org/doc/04-schema.md#replace that way you can tell Composer that it can replace picofeed. Then other dependencies you may have that want picofeed will all know they can use yours.
geoffo joined the channel
# [LewisCowles] a case for vendoring dependencies. Like a fat-jar / package cache
[tantek] and theianjones joined the channel
gRegorLove joined the channel
chrisaldrich, doubleloop1 and TiffanyDoll joined the channel
# Loqi jacky: Zegnat left you a message 10 hours, 13 minutes ago: from yesterday’s call, I think this is what both PHP and Python parsers have tested with as a possible whitespace cleanup alg: https://wiki.zegnat.net/media/textparsing.html
[jgmac1106] and [chrisaldrich] joined the channel
# [LewisCowles] something broke docker. A build normally taking 30-45 minutes has now taken over 2.5 hours
# [LewisCowles] should restarting daemon fix it? Trying that now
# [LewisCowles] ah I've only been pruning containers and images jacky++
[KevinMarks] joined the channel
# [KevinMarks] database-tax << https://twitter.com/timbray/status/1254143366914162688?s=20
# 
@timbray 1/ Dear LazyWeb, need an obscure database recommendation. So, I’m migrating my blog from my 2014 Mac to new 16" Catalina box. I wrote it in 2002 and it’s in Perl. Has a backing database in mysql. However, it’s essentially impossible to use DBD::mysql on Catalina. (twitter.com/_/status/1254143366914162688)
# Loqi ok, I added "https://twitter.com/timbray/status/1254143366914162688?s=20" to the "See Also" section of /database-antipattern https://indieweb.org/wiki/index.php?diff=69640&oldid=69552
[christophe194] joined the channel
# [LewisCowles] perl-tax 😉
# [KevinMarks] resists replying with: well, there's more than one way to do it
# [LewisCowles] just a joke because I never fully got PERL
[snarfed] joined the channel
# [KevinMarks] I did in fact write some perl last year to do something similar - use perl on a Linux box to export data on an mssql database to text files so I can import it into AWS sql
# [KevinMarks] And now we're turning off the windows and Linux box I need to remember how to do it again to get the changes since
themaxdavitt joined the channel
# [LewisCowles] jacky, seems I had to nuke the file-system the repo resided in. Mutative build systems ugh...
gRegorLove_ joined the channel
# [LewisCowles] [KevinMarks] RE: keep all the versions, that is most definitely a dev topic
# [LewisCowles] I support it
# [KevinMarks] Github's big culture change was fork first, ask questions later
# 
[jgmac1106] Thx skenbel.. That one seemed easier... It is difference between embedding and inclusion I don't understand... Wiki people always talk inclusion
# [LewisCowles] last year I came back from holiday to find people in-place updating software, because new was better; but without measurements and buy-in outside of engineering, that is at best posturing
# [LewisCowles] feature-flags and systems configuration are wonderful beasts to enable keeping at least most of the versions
# [LewisCowles] language however seems simplest when mutable
# [LewisCowles] and precise
# [LewisCowles] defining new higher-order publics should not replace or interfere with lower level systems
# sknebel [jgmac1106]: Mediawiki specifically calls it "transclusion" afaik, and I think the user-facing part is IMHO more interesting than any technical aspects. to me, "embed" is more "you embed an entire thing (document, video, Instagram post, ...) as a box on a page", vs wiki transclusion transparently puts things in the page. E.g. the micro-hcard templates on the wiki replace the text inline, it's not obvious that it's from somewhere else
# [LewisCowles] I will challenge that nodejs and just spin up more machines have aged terribly
# [KevinMarks] We're still trying to find good ways to make this behaviour make sense for people - to know when changes will ripple through and when they won't
# [LewisCowles] Sometimes it starts with describing the behaviour, even if in the knowledge it will be revised in future
# [jgmac1106] Ahh okay thx skenbel... Fits the work wikimedia doing to unite all citations into wikidata
# [LewisCowles] I like the ability to track unfurl, transclusion, embedding. I must admit one of those terms did make my eye twitch when first mentioned. It seemed distinctly problematic
# [LewisCowles] outcome based terminology has it's place. I think 1 of the three is solely relegated to silo's
# [LewisCowles] yeah, it's wikipedia hints at some of the problems with it. It feels a lot like someone coined the term to avoid plain language and engage in a bit of chimping
# [LewisCowles] chimping being a polite term for a very specific public display from on-high
BillSeitz joined the channel
# [LewisCowles] I thought I best dfn before another use gets mis-attributed. TBF I should have used the M word, we're all adults AFAIK
billseitz_ joined the channel
# billseitz_ I recently got IndieAuth working for my site again. But when I log in anywhere with it, the remote site shows my ID/name(?) as "Webseitz.fluxent.com wiki FrontPage" rather than "Bill Seitz". How do I fix that? I suspect it's in some combo of rel="me" and h-card. http://webseitz.fluxent.com/wiki/FrontPage
# billseitz_ (custom engine)
# billseitz_ Tried indieweb.org
# billseitz_ and https://webmention.io/
# 
aaronpk right now it's just blank http://webseitz.fluxent.com
# sknebel I.e. I'm a URL too when I edit pages etc: https://indieweb.org/wiki/index.php?title=User:Www.svenknebel.de&action=history
# billseitz_ So, even if I do that, the outcome will be a “username” on these sites that’s my domain, not, say, the name in my h-card
# aaronpk if you want the thing on the wiki where you can use
{{nickname}} then the wiki has a special way to do that which has nothing to do with authentication https://indieweb.org/wikifying#Step_Four:_Add_a_sparkline# billseitz_ gotcha, thx
KartikPrabhu joined the channel
# [KevinMarks] 2 more terms for this there, template and sparkline
KartikPrabhu, billseitz_, geoffo and doubleloop1 joined the channel
# [LewisCowles] one last compile and then sleep
# [LewisCowles] managed to get a working copy of h-feed and export into WeKan export. Need to run it through x-ray
# [LewisCowles] cry at failures etc