#dev 2018-02-23

2018-02-23 UTC
#
[miklb] https://github.com/integrations/slack/issues/316
#
Loqi [bkeepers] #316 404 when clicking link to install GitHub App
#
j12t_ KartikPrabhu: I'm playing with mf2py. Are the nosetests supposed to work out of the box? They seem to access foo.com and bar.org
#
[cleverdevil] Is there a tool out there in indieweb-dev-land that looks at a website and attempts to guess an mf2 structure for it?
#
[cleverdevil] Sort of like an Instapaper-web-service?
#
KartikPrabhu j12t: I have no idea about the tests
#
snarfed [cleverdevil]: https://xray.p3k.io/ ?
#
[cleverdevil] [snarfed] well, XRay assumes that the target site is already marked up with microformats2
#
snarfed ah ok
#
j12t KartikPrabhu: Ah :-) I figured something like it
#
KartikPrabhu j12t: if you want to help write/edit the tests things that would be great. I don't know how that part works at all
#
j12t KartikPrabhu: I'm new to this code, not sure how helpful I can be any time soon on it
#
KartikPrabhu sure, no worries
#
KartikPrabhu eventually it would be good to use some standard testing suite
#
KartikPrabhu not sure what happened on that front
#
snarfed start with running the existing tests :P
#
KartikPrabhu ha sure :P
[kevinmarks] and snarfed joined the channel
#
oodani am i correct in believing that there's currently no way for an e-* property in microformats2 to control which formats it exposes and what it provides for each format? for instance, my site's using markdown - there's no way to expose the un-rendered markdown source as a format in e-content?
#
KartikPrabhu oodani: microformats only cares about the final HTML so no
#
KartikPrabhu what's your use-case for exposing the markdown?
#
oodani it's a more accurate text representation of the content than is returned by textContent, and also, i'd like micropub clients to be able to pull that markdown when i update posts, rather than only the rendered html
#
KartikPrabhu the micropub one makes sense, but other than that I am not sure anyone wants to consume the markdown for your posts
#
oodani that's fair. hmm.
#
KartikPrabhu aaronpk: ^ is there a way currently to expose markdown for micropub editing ?
#
KartikPrabhu oodani: also I don't know of any mircopub editor that uses markdown (?)
#
oodani yeah, i know. one of the nice things about markdown is it still works pretty well without editor support
#
oodani although for longer entries i usually write in vim + marked.app or stackedit :p
#
KartikPrabhu right, but then there is no use-case for it to be in microformats
[eddie] joined the channel
#
[eddie] That would be part of the Micropub source endpoint: https://www.w3.org/TR/micropub/#source-content
#
[eddie] Your endpoint can return whatever text content it wants in the content attribute. Whatever gets returned from source would be available in the Micropub editor.
#
KartikPrabhu one use-case I can see is if you use markdown to edit but store your posts as html+mf2
#
oodani ah. the source content doesn't need to match the entry's "real" content?
#
[eddie] Nope, it matches the “editing” form of your content
#
KartikPrabhu oh ok that makes sense
#
[eddie] For people that edit HTML (like tantek) it would be HTML, for people who edit in markdown (like me) it returns markdown
#
KartikPrabhu is way behind on micrpub spec
#
oodani and yeah, there's no micropub clients that support markdown yet, but there's also no way to express explicitly' that an e-content value is markdown yet :o seems like editors will probably start adding that support if markdown e-content starts popping up in the wild
#
oodani sounds like tweaking my source endpoint will probably do the trick in any case. currently it's literally just parsing the requested page with mf2py, so it's identical to the output format :p
gobengo, snarfed and renem joined the channel
#
@bengo @scott_gruber @w3c @grabaperch Nice! Are you using https://webmention.io/ to receive the mentions, or coding a perch plugin? The more implementations, the better! (twitter.com/_/status/966865540856729600)
radicaledward101 and snarfed joined the channel
#
aaronpk Looks like that question was covered
#
aaronpk The query is meant to return the "editable" version of your content, whatever that means to you. If you return text, clients will treat it as text, butthat can be markdown text if you're willing to edit markdown like you edit plain text, which is what markdown is by design so it should be fine
#
KartikPrabhu yeah
#
KartikPrabhu was behind on new micropub stuf
[xavierroy] joined the channel
#
j12t Speaking of testing ... I just came across this behavior of the mf2py library: https://github.com/kartikprabhu/mf2py/issues/60. It would be very nice to have test cases for things like that, nicely annotated with explanations why this is the correct (or not correct) behavior for the library. Just saying ...
#
Loqi [jernst] #60 Repeated fields
#
KartikPrabhu j12t: just replied to your issue. I suspect it is because mf2py does not handle a mixture of mf1 and mf2 cleverly yet
#
KartikPrabhu but your example does not seem to be too bad for consuming
#
j12t Ah. Happy to provide HTML, but where to put it? We should be turning this into automated test cases over time.
#
KartikPrabhu j12t: you can put it in the github issue for now so I can see why it is doing that
#
j12t (Personal pet peeve: I hate doing things twice). So I put the files here: https://github.com/jernst/mf2py/tree/master/test/future-examples
#
j12t Ideally, if all that changed was the theme, the uf content should not change IMHO.
#
KartikPrabhu j12t: wordpress themese do weird stuff
#
j12t Well, assuming that the two themes show the same post(s) etc.
#
KartikPrabhu wordpress themes can change classes so they can insert different microformats
#
j12t Why would they want to? It seems to me the microformats represent the semantics of the displayed content, and those are the same regardless of whether the theme is colorful or boring.
#
KartikPrabhu I am not sure why they "want" to, but I have seen them do that
tantek joined the channel
#
j12t On this subject ... some thoughts on automated testing. Blogged here: https://upon2020.com/blog/2018/02/automated-testing-for-the-indie-web/
#
KartikPrabhu sounds complicated
#
snarfed testing++
#
Loqi testing has 4 karma in this channel (23 overall)
#
aaronpk I keep getting tempted to write a test suite for the Microformats for all the different kinds of posts on my site so I don't inadvertently break something when I make changes to my html
#
aaronpk Thankfully it sounds way easier to do on my site than this blog post describes for wordpress 😂
KartikPrabhu, gobengo and [kevinmarks] joined the channel
#
[kevinmarks] We have a test suite already, so we should make sure we are using it
#
[kevinmarks] https://github.com/microformats/tests
#
Loqi [microformats] tests: Microformats test suite
cweiske, loicm, schmarty, petermolnar, grantcodes, chimo and jeremycherfas joined the channel
#
sknebel yeah, although the infrastructure around it isn't great, and it hasn't been updated in a while it seems
#
sknebel (and I personally am not a fan of the structure by type)
myfreeweb, plindner, eli_oat[m], zoglesby, grantcodes, schmarty and mindB joined the channel
#
@clivewalker @bengo @scott_gruber @w3c @grabaperch Perch supports webmentions in core. See https://docs.grabaperch.com/addons/blog/webmentions/ and https://grabaperch.com/blog/archive/2017-07-26-displaying-a-webmentions-facepile (twitter.com/_/status/966967702836404224)
[kevinmarks] joined the channel
#
[kevinmarks] Fixing the infrastructure is a good idea.
barpthewire joined the channel
#
sknebel Biggest thing for that probably would be making sure all parsers have either conversion tools, or a standardized way of doing testing over HTTP or a command line interface or ...
#
sknebel And of course reviewing what kind of tests are missing/incorrect
leg joined the channel
#
aaronpk I'm not talking about parser level problems, I'm talking about vocab things and testing whether my html keeps mf2 classes in the right place when I make changes
Loqi and eli_oat joined the channel
#
loqi.me edited /moderation (+11) "sknebel added "[[spam]]" to "See Also"" (view diff)
#
@scott_gruber @clivewalker @bengo @w3c @grabaperch Hmm. Been going over those and in Perch I see at least one real webmention (there is 7 in thr control panel) and approved it, but nothing shows up yet a blog post page...yet! (twitter.com/_/status/967046348745850880)
#
@clivewalker @scott_gruber I don't display them with a facepile but it's on my site now https://www.clivewalker.me/blog/2017-07-18-exploring-webmentions (twitter.com/_/status/967052723140128768)
snarfed, KartikPrabhu, barpthewire1 and singpolyma joined the channel
#
singpolyma Does it make sense to use h-cite without a title, for h-entry replies, like so: http://lpaste.net/3978319217194696704 ?
#
aaronpk i think in general h-cite should match the thing you're citing
#
aaronpk but until the new implied name rules are finalized and implemented, parsers will generate a name property for that anyway
#
aaronpk https://pin13.net/mf2/?id=20180223154315635
#
singpolyma Right, so they'll end up thinking it has this garbage title
#
singpolyma hmm, ok
#
singpolyma so maybe not worth it
#
snarfed hey singpolyma, long time no see!
#
sknebel So you want to link to the reply, but not show it's content?
#
singpolyma sknebel: correct
#
Zegnat An h-cite of an untitled thing might be a little weird, yeah. If you are pointing to a resource (a comment at an external URL) and also want to give that resource’s author information, an h-cite still seems right to me.
#
Zegnat I’d do class="u-comment h-cite" though, as per https://indieweb.org/comments#How_to_markup
#
singpolyma it's not a comment, it's a post I'm replying to in this case
#
Zegnat Ooh, its the reply context, sorry. (Then add a u-in-reply-to? :P)
#
Zegnat You’ll probably have to add an explicit empty name element though :(
barpthewire, snarfed1, snarfed, jackjamieson, aaronpk, tantek and [cleverdevil] joined the channel
#
[cleverdevil] Linode issues, [eddie]
#
aaronpk back now
#
aaronpk https://status.linode.com/incidents/2088k0rt4rb9
[eddie] joined the channel
#
[eddie] ahh makes sense
#
[eddie] lol literally the only p3k server I rely on got me. haha
#
[eddie] atlas
#
[eddie] 😆
#
Loqi ahahahaha
#
aaronpk the good/bad thing about connecting to IRC from the same server as my website is I notice when my website is down immediately
#
[eddie] ohhh wow, that is somewhat helpful
#
[eddie] Obviously it restricts your communication lol
#
[eddie] but the awareness is helpful
#
[eddie] I would probably not notice my site was down until I tried to do a micropub post to it
#
[eddie] or I tried to use microsub
#
[eddie] Hmmm aaronpk no rush, but that IndieNews error is still happening, even though I no longer have doubled u-syndications: I’m sure it’s something wrong on my end but I can’t figure out what would cause it to 500 out: https://telegraph.p3k.io/webmention/12xmulNtxFMNOB8Tcx/details
#
aaronpk oh weird
#
aaronpk "still" happening?
#
[eddie] https://chat.indieweb.org/dev/2018-01-14/1515915640199200
#
Loqi [[eddie]] !tell aaronpk: I think indienews is broken. 500 error on webmention? https://telegraph.p3k.io/webmention/12BCWO1g1Rvu0uSdYj/details
#
aaronpk whoops, missed that
#
aaronpk probably shoulda filed an issue for that :)
#
[eddie] I thought it was me
#
[eddie] at the time you mentioned the double u-syndication
#
aaronpk oh
#
[eddie] I just fixed that double syndication today and found out it wasn’t that 🙂
#
[eddie] I can file an issue for it then :thumbsup:
loicm joined the channel
#
aaronpk huh my local indienews can parse it fine
#
staceydepolo.com edited /GoDaddy (+394) "/* Progress */ added link to new post on GoDaddy's blog. Fixed references to authors." (view diff)
#
tantek wonders if the newsletter was affected
#
tantek snarfed++ thanks for updating the Authorize bridgy prompt for publish to GitHub!
#
Loqi snarfed has 43 karma in this channel (354 overall)
#
snarfed welcome!
#
tantek noticed you got reacji working too!
#
snarfed yes!
#
tantek did you document which precise emoji chars are supported (or preferred?) for the different GitHub reactions?
#
tantek hmm maybe this should be documented on /GitHub at least
#
tantek hah it's mentioned here https://indieweb.org/GitHub#Features
#
snarfed yup. right now just the exact chars they support. https://brid.gy/about#github-reaction
#
snarfed your forgiving proposal w/similar chars was good, i may do that eventually
#
tantek.com edited /GitHub (+152) "/* Features */ which reacji and text equivalents" (view diff)
#
aaronpk oh hey, htmlpurifier finally published a release!
#
aaronpk (this is related to your bug eddie I swear ;-)
#
tantek.com edited /GitHub (+10) "/* Bridgy Publish */ reacji too" (view diff)
#
snarfed.org edited /GitHub (-285) "bridgy backfeed, prompt" (view diff)
#
tantek snarfed, did you capture the forgiving proposal in an issue? just curious where you're tracking it
#
snarfed tantek: no issue, just my todo list. feel free to file an issue if you want!
#
tantek it's not super critical, just a nice usability touch
#
tantek ok
#
snarfed right
#
tantek has anyone built-in progressive back-off retrying for webmention receiving (and retrieving the source)?
#
tantek from #indieweb. e.g. if you try retrieving the source and get a server error (5xx), perhaps retry in 5 min, then an hour, then a day, then give up?
#
tantek.com edited /GitHub (+519) "note old screenshot, solution, and update" (view diff)
#
aaronpk hm that wasn't where I thought that was going
#
aaronpk if you receive a webmention, it's very likely the sender is still online
#
aaronpk I was expecting that question in regards to sending webmentions
#
tantek that too! (much more likely)
#
tantek right now I don't retry at all (because my webmention sending is synchronous with publishing lol)
#
aaronpk I also don't have any retry logic
#
tantek do you keep any metrics on 5xx errors?
#
tantek I suppose you're more likely to run into it at the endpoint discovery step?
#
aaronpk I do!
#
tantek though if a site uses an external service (wmio etc.), then you'd get it at the send step (if the proxy was down)
#
aaronpk ah, my discovery logging doesn't have that info
#
tantek so three steps in the protocol such retrying could apply to: discovery, sending, verification
#
tantek is there existing brainstorming on this?
#
aaronpk not much https://indieweb.org/Webmention-brainstorming#retrying.2Ferror_handling
#
aaronpk I do have retry logic for fetching feeds for my reader
#
tantek what is retry
#
Loqi It looks like we don't have a page for "retry" yet. Would you like to create it? (Or just say "retry is ____", a sentence describing the term)
#
tantek well let's start capturing examples then, and if there are any patterns, perhaps we can documenting guidelines
#
snarfed yes, bridgy does progressive backoff retry. exponential, starting at 30s, capped at 1h iirc, and it gives up after 24h
#
snarfed this is somewhere on the wiki, not sure where
#
aaronpk that link I found
#
snarfed cool thx
#
tantek anyone else notice that GitHub's reacji documentation has a screenshot with an emoji that doesn't show up in GitHub's current reacji UI? https://help.github.com/articles/about-conversations-on-github/#reacting-to-ideas-in-comments
#
tantek 😢
#
tantek wondering if that's a hidden feature
#
tantek snarfed, does the GitHub reaction API use an enum (1...6) or emoji?
#
aaronpk [eddie]: try publishing again!
#
snarfed tantek: i don't see that emoji in either of their apis. https://developer.github.com/v3/reactions/#reaction-types , https://developer.github.com/v4/enum/reactioncontent/
#
tantek but it's in their docs
#
snarfed doc is probably out of date
#
[eddie] aaronpk: Accepted!
#
aaronpk woo
#
aaronpk It was the mailto link
#
[eddie] oh! strange
#
[eddie] lol
#
[eddie] yay! I’m IndieNews ready again! haha
#
Loqi nice
#
Loqi does a happy dance!
#
[eddie] Thanks loqi!
#
Loqi you're welcome
#
[eddie] loqi++
#
Loqi loqi has 7 karma in this channel (448 overall)
tantek and snarfed joined the channel
#
j12t kevinmarks: I didn't know about those microformat tests. What code exactly does this test suite test?
#
aaronpk it tests the parsers themselves
#
aaronpk not what youre after afaict
#
j12t nm, I think I get it now. Is anybody running these tests against anything on a regular basis?
#
j12t btw, it's great to have those!
snarfed joined the channel
#
bear I'm more than happy to talk about uptime, 4 9's and other requirements for them if anyone is interested
#
snarfed bear++
#
Loqi bear has 21 karma in this channel (197 overall)
#
snarfed hey bear will you also port my projects from circle 1 to 2 for a reasonable fee? 😂😭
#
KartikPrabhu lol
#
bear :)
#
bear probably - I know I have to migrate my own
#
tantek.com edited /GitHub (+464) "/* OAuth UI */ embed new screenshot" (view diff)
#
bear the movement from 2 9's to 3 9's is easy enough that people then assume the transition from 3 to 4 9's will also be easy
#
snarfed heh good point
#
bear but then they find out that 4 9's also requires attention to the server and it's environment as well as the software stack
#
bear 4 9's is when you move from "r"edundancy to "R"edundancy and that it has to be actively managed
#
bear IMO the best way to get 4 9's is to fake it - cache your outer edge so your customers don't notice - but then you have to deal with writes
#
tantek hard to fake for community editable sites
#
tantek or rather, caching won't fake that
#
bear yep
#
tantek not a bad fallback tho
#
bear 99.9 is Good Enough(tm) for the vast majority of anything that is community driven
#
snarfed bear++
#
snarfed arguably even the vast majority, period :P
#
loqi.me created /Lua (+196) "prompted by tantek and dfn added by bear" (view diff)
#
bear snarfed++ indeed
#
Loqi snarfed has 44 karma in this channel (355 overall)
#
bear 4 9's will cost you serious money - even as a hobbyist
#
bear 5 9's is very rare and very expensive - i've only worked a handful of jobs where 5 9's was the goal
#
tantek and if it doesn't, as a hobbyist, then whatever you figured out likely has a business model
#
tantek (re: 4 9's costing serious money)
#
bear yep
#
bear that's why community specific CDN vendors are so popular
#
tantek though I wonder if such qos guarantees have exceptions for levels of adversaries
#
tantek e.g. 4 9's except if for IoT bot army DDOS
#
tantek or except if you happen to piss off a state-level-actor
#
tantek etc.
#
bear when you track 9's you spend a lot of time explaining why you missed your targets
#
tantek many of these high 9's don't seem too hard to achieve except for specific failure modes
#
bear yep - we have had to make certain business decisions for what level of DDoS we can handle and have let our board know at what point we will just fall over
#
tantek so then it seems like a question of solving those rather than # of 9s
#
bear not really - the devil-in-the-detail is where do you measure uptime
#
bear if it's customer facing web site that is 99.99 then you have to ensure that everything behind it is at 99.995 or better
#
snarfed at enough 9s - usually over 3 - pretty much all remaining downtime is "specific failure modes"
#
snarfed ie, if you're succeeding, every new downtime is a fundamentally new *class* of failure
#
bear yes
[cleverdevil] joined the channel
#
[cleverdevil] ^^^
#
snarfed put another way, https://snarfed.org/2018-01-03_27966
#
Loqi [Ryan Barrett] We are in the business of novel outages.    – Google Cloud SRE …since ideally you find and fix each outage’s root cause. Admirable. Definitely the right idea!
#
bear that's why we (internally) measure uptime in weekly cycles
#
snarfed thankfully we (indieweb) shouldn't really worry about any of this right now
#
snarfed "three 9s should be enough for anyone"
#
tantek right, hence my point about explicitly specifying fallback/retrying in our protocols accordingly
#
tantek so that newer implementers start with implementing in such a way that is sufficiently forgiving of "only" 3 9s indieweb sites
#
tantek instead of falling the way of SMTP which has become an oligopoly of high volume centralized services :(
#
KartikPrabhu heh oligopoly is a funny word
#
KartikPrabhu funny sounding
#
snarfed definitely good to explicitly talk about "three 9s are enough." i wouldn't use it to drive retry/fallback though. ime more failures are either shorter (network, reboot, etc) or longer (bugs) than hosting provider outages.
jackjamieson joined the channel
#
bear yes, IMO the only place retry/fallback should be raised is if someone is implementing a community service like brid.gy
#
bear as they will need to be polite denizens in the IndieWeb community
#
aaronpk I think there's a case for some best practices around retrying when sending webmentions in general
#
snarfed huh, really? retry seems appropriate for eg individual sites sending webmentions too
#
snarfed heh
#
bear from the point of view of receiving webmentions - what is the difference between someone trying 10 times in a row or 10 people trying once?
#
bear I guess maybe have a "don't be rude" mention - but retry/fallback seems like overkill
#
aaronpk I don't follow
#
bear you said there is a case for best practices
#
aaronpk if I try to send a webmention to someone and get back a 500 server error or time out on the discovery, I would think I should try that again at some point in the future
#
snarfed the goal isn't really to avoid DOSes, it's to improve the chance of getting through successfully
#
bear oh - I was thinking of a different use-case
#
bear I see your point
#
bear enables his failback handler and waits to reply
#
tantek what snarfed says
#
bear yes, for some reason I was combining the two conversations in my head
#
tantek bear, though yes, that's the other side of the retry conversation
#
bear (the peril of trying to track this conversation while in a meeting - I should know better)
#
tantek retry enough to be forgiving to 3 9s (or less?) indieweb sites, but don't retry so much that you cause DOS
#
tantek both points are valid and worthy inputs to figuring out a good retry convention / policy / best practices
chrisaldrich joined the channel
#
www.boffosocko.com uploaded /Special:Log/upload "uploaded "[[File:HWC Los Angeles 2018-02-23.jpg]]" https://indieweb.org/File:HWC Los Angeles 2018-02-23.jpg"
#
www.boffosocko.com edited /Perch (+455) "Clive Walker Example; Articles on Webmentions and Facepiles" (view diff)
#
www.boffosocko.com edited /Anchor (+287) "criticisms" (view diff)
KartikPrabhu and abbot joined the channel
#
mifga any folks in indieweb use Tinderbox from Eastgate?
snarfed joined the channel
#
KartikPrabhu what is tinderbox?
#
Loqi It looks like we don't have a page for "tinderbox" yet. Would you like to create it? (Or just say "tinderbox is ____", a sentence describing the term)
#
KartikPrabhu what is eastgate?
#
Loqi It looks like we don't have a page for "eastgate" yet. Would you like to create it? (Or just say "eastgate is ____", a sentence describing the term)
#
KartikPrabhu looks like no
[cleverdevil] and [eddie] joined the channel
#
aaronpk thanks twitter "Exception: EOFError"
#
tantek whoa
#
tantek end of feed error? ;)
#
[cleverdevil] lolwut.
#
aaronpk something with the streaming api