2014-11-24 UTC
# 00:04 gRegor` Domain Name Scarers
parzzix joined the channel
# 00:10 aaronpk fiatjaf: interesting idea about the third-party webmention collection endpoint. I don't really see how it would actually work though
# 00:11 aaronpk because at some point you'd have to tell someone "hey go to this site to see all the webmentions you've been getting"
# 00:11 aaronpk and if there's one of these services, why wouldn't there be three of them?
# 00:11 aaronpk at which point people would have to check a bunch of places and each would have some subset of the mentions
# 00:12 aaronpk this also relies on webmention senders to hard-code one of these services as a fallback
j12t joined the channel
# 00:17 aaronpk if you don't accept webmentions and i've been linking to your site, I actually publish a page where you can see all the posts I've mentioned you in
# 00:20 aaronpk (it's kinda fun to plug in random domains there and see what happens!)
# 00:20 gRegor` Ooh, I didn't know you had that, aaronpk
# 00:20 aaronpk it's relatively recent, only been there a month or so
# 00:20 aaronpk and i haven't figured out how to surface links to those pages elsewhere on my site
# 00:21 gRegor` "in reply to [domain] (all links to [domain]" maybe, for replies at least.
# 00:22 gRegor` Just realized that
# 00:22 aaronpk I make a point of not having any sort of third-person text
# 00:23 aaronpk the thing should be obvious when looking at it on the screen without explanatory text. that's why i spent so much time making my reply context look right.
# 00:25 gRegor` Something in the CSS seems to make it 48 x 50 instead of square
ricmac joined the channel
# 00:26 gRegor` Ah, the container is 48x48, but the .minicard .photo is height: 50
fmarier joined the channel
fmarier joined the channel
# 00:35 aaronpk don't we have some screenshots somewhere that show when twitter credited the app a tweet was posted with?
# 00:36 gRegor` Hehe, might not have noticed with someone else's photo. :)
# 00:37 aaronpk darn, was removed just in 2012. I wonder if I can fake a screenshot for illustrative purposes
fmarier joined the channel
# 00:45 aaronpk (that one's real btw. dug it up from my flickr account)
Pea1 joined the channel
KartikPrabhu joined the channel
chrissaad and mlncn joined the channel
# 01:57 ben_thatmust seems really weird having it parse as category then h-card as a sub-element, so i'm not sure if i'm doing something wrong
parzzix, mdik_, mlncn, j12t and KevinMarks__ joined the channel
gr0k and demis_ joined the channel
amblin joined the channel
fmarier joined the channel
# 04:03 GWG KartikPrabhu: Isn't that the truth?
# 04:04 aaronpk I'm making a webdav -> flickr bridge right now. Next up is repurposing the code for webdav -> micropub
# 04:06 GWG I'll look forward to the demo. You do put out interesting thigns
snarfed1 joined the channel
nonboong__ joined the channel
# 04:17 aaronpk lol at using skitch to take screenshots of the skitch setup process
j12t joined the channel
# 04:18 Loqi Skitch is an application for OSX which can take screenshots, add annotations, and share the images to multiple destinations such as Evernote, FTP, or a WebDAV endpoint http://indiewebcamp.com/Skitch
npdoty joined the channel
# 04:36 kylewm ben_thatmust: I think I missed the discussion, why is the person tag "p-category h-card"?
# 04:36 GWG kylewm: Because it is a tagging of a person?
# 04:37 GWG kylewm: I usually go with the flow myself.
# 04:38 kylewm is grepping the logs like he should have in the first place
# 04:43 GWG Of course, I'm usually confused about these things.
fmarier, Electrin, parzzix, j12t, snarfed, Haxxa, npdoty and snarfed1 joined the channel
j12t, chrissaad, wolftune, loic_m and cweiske joined the channel
j12t joined the channel
npdoty, Jihaisse, cweiske and j12t joined the channel
krendil, KevinMarks, friedcell, petermolnar, acegiak, adactio, Sebastien-L, Jihaisse, stream7, eschnou, j12t, cweiske, alanpearce, Electrin and Pierre-O joined the channel
# 12:03 Loqi Ok, I'll tell him that when I see him next
j12t, eschnou and friedcell joined the channel
# 12:26 fiatjaf !tell aaronpk http://aaronparecki.com/links-to is the same as what a third party service would do, but done in your own site. if there are people outsourcing their webmention endpoints they could possibly outsource the 'links-to' service also. for people wanting to check webmentions of them, to check every website of the world for a 'links-to' section seems at least as difficult as checking a handful of webmention-storage-services.
# 12:26 Loqi Ok, I'll tell them that when I see them next
# 12:28 fiatjaf !tell aaronpk: anyway, it is cool that you're doing it by your own. this is probably the best thing to be done (well, we still have the problem of letting the mentioned people know and grab their webmentions, but who cares).
# 12:28 Loqi Ok, I'll tell them that when I see them next
friedcell, voxpelli, CaptainCalliope, b0bg0d_____, jden, benward_____, JonathanNeal, Garbee, tommorris, jancborchardt, sivoais, bret, mattl_ and hadleybeeman_ joined the channel
ben_thatmust_ joined the channel
sivoais, Sebastien-L, wolftune, j12t, verdi_, shiflett, cweiske, veselosky, parzzix, brianloveswords, demis, mlncn, Pierre-O, danlyke_, gr0k, snarfed, mlncn-agaric, friedcell, PMurphs, mko, gRegor`, chrissaad, richardlitt and npdoty joined the channel
# 17:05 Loqi aaronpk: fiatjaf left you a message 4 hours, 38 minutes ago: http://aaronparecki.com/links-to is the same as what a third party service would do, but done in your own site. if there are people outsourcing their webmention endpoints they could possibly outsource the 'links-to' service also. for people wanting to check webmentions of them, to check every website of the world for a 'links-to' section seems at least as difficult as checking a handful of webmention-storage-services.
# 17:05 Loqi aaronpk: fiatjaf left you a message 4 hours, 36 minutes ago: anyway, it is cool that you're doing it by your own. this is probably the best thing to be done (well, we still have the problem of letting the mentioned people know and grab their webmentions, but who cares).
# 17:05 mko Oh, that's nice. I like that "links to" page.
# 17:06 mko aaronpk: How are you doing that analysis? Just scraping the links out and aggregating by domain?
# 17:06 aaronpk whenever a post is saved, it parses the content for links and adds the domains to a database table
tilgovi, j12t, Pea1, Pierre-O, snarfed, friedcell, ricmac, gRegor`, shiflett and KevinMarks joined the channel
# 18:26 snarfed current itch: de-duping native indie comments (likes, RTs, etc) and their silo POSSE copies
KevinMarks and j12t joined the channel
# 18:32 kylewm snarfed: oh lordy, me too ... i have the wikipedia page for Levenstein distance open in my browser at home
# 18:33 snarfed or maybe stripping trailing … (if any) and then prefix match?
# 18:34 snarfed regardless. more importantly: glad to hear it's itching you too :P
# 18:34 kylewm yeah if i trimmed the @name and did prefix match, that would work most of the time
# 18:38 gRegor` I think I would check for syndication URLs on the received wms and compare against the URLs of the other wms
# 18:39 kylewm gRegor`: that's exactly what i used to do, but lots of people syndicate and add syndication urls after sending webmentions
gr0k joined the channel
# 18:51 snarfed hmm. kylewm you may have a point. the text may often be the same, or a prefix, but the html tags and formatting may often differ
# 18:51 snarfed maybe stripping tags and ignoring whitespace in the comparison would be enough
gr0k, richardlitt and squeakytoy joined the channel
ricmac and KartikPrabhu joined the channel
ricmac_ and tantek joined the channel
# 19:24 Loqi tantek: ben_thatmust left you a message 7 hours, 21 minutes ago: does this seem to be correct in its parsing? the category and h-card ending up nested seems odd, but looks to be correct for the algo. Want to double check it is correct before i build on it. https://gist.github.com/dissolve/2911540317b48514f4c6
# 19:25 tantek ben_thatmustbeme: great test case, it's an interesting challenge of both implying property names and minimal parent/child microformats objects
# 19:27 tantek here's a general principle - the same markup for a property should not be causing that property to occur in *both* a microformat and one embedded inside - such a property should only be showing up on one.
# 19:27 tantek thus it's up to the parsing algorithm to be tightened up to prevent that
# 19:27 GWG aaronpk, I keep feeling like there are uncharted parts of your site I should be studying
# 19:28 aaronpk I haven't figured out how to link to everything properly yet
# 19:29 gRegor` what is spidering?
# 19:31 gRegor` might do that on my site someday
# 19:33 tantek for anyone interested in iterating on /microformats2 parsing iterations to fix the example ben_thatmustbeme pointed out - I'll be chatting it up in #microformats
richardlitt joined the channel
# 19:58 tantek reads-up about de-duping between kylewm, snarfed, gRegor`
# 19:59 tantek kylewm: "lots of people syndicate and add syndication urls after sending webmentions" - hmm, we can at least provide implementation guidance to do that in a better order.
eschnou joined the channel
richardlitt joined the channel
# 20:04 tantek so that's easy - always (re)send a webmention after you add the syndication URL, even if that means you send more than one
# 20:08 kylewm I agree with that, but would still like some more relaxed matching, since some wordpress users still publish u-syndication inconsistently or not at all
# 20:09 tantek kylewm - no disagreement, just trying to reduce the frequency that fuzzy matching is needed in practice.
# 20:09 tantek having a deterministic path to de-duplication when publisher and receiver are following the protocols is a good thing.
# 20:10 Loqi tantek meant to say: having a deterministic & dependable path to de-duplication when publisher and receiver are following the protocols is a good thing.
ricmac, richardlitt, Pea2, KartikPrabhu and krendil joined the channel
KartikPrabhu and caseorganic joined the channel
Sebastien-L and shiflett joined the channel
Pea1, tantek and pauloppenheim joined the channel
gr0k joined the channel
wolftune joined the channel
ricmac joined the channel
# 21:26 Loqi Ok, I'll tell them that when I see them next
# 21:27 Loqi Ok, I'll tell them that when I see them next
# 21:28 bret aaronpk, dietrich either of you in town Dec 3rd?
# 21:29 bret school homework and presentations will be over with by Wed... Freedom!
ricmac joined the channel
# 21:35 tantek we have a bunch of Mozilla groups gathering to do an in-person work week
KevinMarks joined the channel
ricmac, lukebrooker and tantek joined the channel
fmarier joined the channel
ricmac and fmarier joined the channel
# 22:31 bret Is it still POSSE if my syndication handler finished before my site publishes the new post? >;)
# 22:32 bret its a little bit easier to get the syndication data into my post file before publishing
# 22:32 aaronpk maybe find a way to build a single page of the site instead of building the whole site each time?
# 22:33 tantek I do this order: 1) live on personal site, 2) PuSH notifications, 3) Twitter POSSE, 4) Send webmentions
# 22:33 bret I just dont have file updating implemented yet
# 22:36 aaronpk tantek: when in the sequence do you update your post with the POSSE URLs?
# 22:37 tantek aaronpk - as part of 3) Twitter POSSE. When the API call returns with the tweet id, that gets saved back to Falcon storage.
# 22:38 aaronpk i forgot about PuSH. Do you have any idea how many consumers you have?
# 22:38 bret its the same application/event creating the Twitter posse and personal website post. so im just gonna go ahead and say it counts
# 22:38 tantek aaronpk - I have no idea since I delegate to Google's hub and they handle that
# 22:39 aaronpk does superfeedr tell you that kind of thing? That'd be a good reason to use superfeedr instead of the google one
# 22:39 tantek I know evanp used to subscribe from statusnet
# 22:39 aaronpk seems like something people (and/or brands) would pay for
# 22:39 tantek seems like something we should simplify so we don't need to use 3rd party hubs / bottlenecks
# 22:40 bret does anyone see an issue with rephrasing "syndicated to twitter/instagram"/"also on twitter/instagram" to "reply on twitter/instagram/your own site"?
# 22:40 tantek we just haven't gotten around to simplifying PuSH yet becuase not enough folks are using it to build a reader
# 22:40 aaronpk why? I'm a fan of making things possible to spin off as separate services if desired
# 22:40 tantek or rather, barnabywalters's is still getting charged for using superfeedr :/
# 22:40 tantek I'm a fan of making things all indie implementable if desired
# 22:41 tantek and that's more important than enabling more bottlenecks
# 22:41 aaronpk *if desired* is the key tho. don't want to require things be run individually
# 22:41 tantek anything that can be indie implemented can also be done as a service
# 22:41 tantek so I don't worry about the "as a service" bit
# 22:42 aaronpk try delegating your webfinger to a service. that sounds hard.
# 22:42 tantek webfinger is not a feature, it's obsolete plumbing
# 22:42 aaronpk can do it on apache/nginx with some clever rewrites
# 22:42 aaronpk bret: sure you can implement it directly. my point is you can't delegate it to a service
# 22:43 tantek aaronpk totally unclear. why bring up webfinger at all outside of a use-case?
# 22:43 aaronpk webfinger can't be delegated to a service, so nobody is using it
# 22:43 aaronpk you said "anything that can be indie implemented" and I wanted to point out a counterexample of somethign that also failed
# 22:43 Loqi Homebrew Website Club PDX on Wednesday, Dec 3, 6:30pm at Esri Portland R&D Center
# 22:44 tantek I suppose I'm making certain design assumptions then for "indie implemented"
# 22:44 tantek which means my generalization doesn't hold. you're right.
# 22:45 aaronpk bret: i dunno. this conversation isn't actually about webfinger tho ;)
# 22:45 tantek bret, aaronpk did calagator ever mark-up their events with h-event?
# 22:45 tantek and are they setup to receive webmention RSVPs yet?
# 22:47 bret be aware, webfist was an attempt to delegate webfinger
# 22:47 tantek bret - do you use/support webfist on your own site?
# 22:48 aaronpk MAN... i really need an interface for creating events
# 22:48 aaronpk basically I want facebook's event creation interface as a micropub client
# 22:48 bret i had a webfinger profile set up on my own site for a bit though, not sure if its still working]
# 22:52 aaronpk do you have documentation of the process of creating an event on FB?
# 22:55 tantek I'm not particularly happy with *anyone's* UI for creating events
# 22:55 aaronpk maybe the new upcoming will be a fresh take on it
# 22:56 aaronpk that's why facebook's is nice! they require very little, and slowly add fields to the form as you fill them out
# 22:56 tantek seems like all the attempts to expand on natural language event creation have died
# 22:56 tantek some work went into google calendar, and some into siri
# 22:56 aaronpk can confirm. was able to create an FB event *only* by typing "test" in the name field
# 22:57 tantek like, what would be the 'tweet' equivalent of creating an event?
# 22:57 tantek that kind of prompting / wizard style is ok but still annoying
# 22:57 tantek because it has no idea what details to get next
# 22:57 tantek doesn't calagator have a good event creation UI/
# 22:57 aaronpk and the "add a time" part only asks for a starting time.
# 22:58 aaronpk then it says "end time?" after you add the start time
# 22:58 tantek of course nearly none of them let you simply paste a URL and go grab the h-event from it :P
# 22:58 tantek most events already exist somwhere on the web
# 22:58 aaronpk "Import one or more events from a feed or web page in one of the following formats: Facebook, hCalendar, iCalendar, Meetup, and Plancast"
# 22:58 tantek were you able to paste the IWC event link into Calagator and have it "just work"
# 23:03 bret natural language event creation never seems to do what I want, google or apple
# 23:04 bret seems reasonable to get calagator to create events based on an h-event permalink
# 23:04 aaronpk especially now that calagator *publishes* h-events
# 23:07 bret gh repo issues should have silent +1 buttons (if the owner allows it)
# 23:07 aaronpk the "tweet equivalent" of posting an event sounds like just posting a note saying "drinks at bailey's at 5pm" with no structured data
# 23:08 aaronpk I find it interesting that they added that feature with no ability to add structured data (you can @-mention venues though) and no actual RSVP mechanism, just an "i'm interested" response and comments
# 23:09 tantek it's the exact same problem as auto-completing people and venues
# 23:09 tantek it shouldn't NLP *after* the fact (i.e. what google/apple do)
# 23:09 tantek but rather auto-complete *as you're typing* the way FB does with names
# 23:11 aaronpk now you're giving me ideas for an event creation interface for Quill
# 23:11 tantek seems like this could all be wrapped into a single utility function you could call as people were typing a word
# 23:12 aaronpk the other big challenge with events is getting the address and lat/lng of the venue
# 23:12 tantek auto_complete_suggest(string, datatype, existing dataset) and it returns a list of suggestions
KartikPrabhu joined the channel
# 23:12 aaronpk unfortunately venues don't always have websites or if they do, don't have an h-card
# 23:13 tantek aaronpk - on your own site, venues always have a URL
# 23:14 aaronpk and even then, what if the venue doesn't exist anywhere yet?
# 23:14 tantek the 99% case is venues already exist with a URL
# 23:15 tantek is this because you don't post venues on your own site?
# 23:15 aaronpk same thing, i would have had to enter the data on my site at some point
# 23:16 tantek presumably you've checked into that office before :P
# 23:16 tantek and in 6 months you've been there more than 99 times
# 23:16 tommorris tantek: mine is virtually nonexistent because the API is dead.
# 23:20 tantek aaronpk - anyway, I'd just start with an export of all my Foursquare checkins
# 23:20 tantek heck, I still have my old dodgeball export too
# 23:20 tantek import all the venues in those to your own site
# 23:20 aaronpk i look forward to exporting all my foursquare checkins
# 23:21 aaronpk sadly I think that's gonna have to wait until after I refactor post types
# 23:21 tommorris tantek: CloudMade - was using it for doing the OSM lookups. I was hoping to replace it with my own search server before they killed it but life and work got in the way.
# 23:22 tantek tommorris: is an API to go from lat/long (from Web GeoLocation API) to list of venues?
# 23:22 tommorris tantek: CloudMade used to provide an API you could call that you could give <x,y> to and it'd return you a list of N nearby venues from OSM. then they decided that they only wanted to deal with "enterprise" customers, by which they mean rich bastards.
# 23:24 tommorris you can call the OSM API (or a few of the read-only mirrors) but the OSM API is harder to use because you can't get it to give you back a filtered nearest-N-POIs-to-location. All you can do is grab a box and get everything in it.
# 23:25 aaronpk i thought there was at least a permissive license?
chrissaad1 and ricmac joined the channel
# 23:26 tommorris the OSM API problem: if you are in the middle of the Arizona desert or out in the Scottish highlands, you need a much bigger box than if you are on Oxford Street or on the corner of Market and Castro. :)
# 23:26 tommorris but you can't know how big the box is until you get it back and it's too small. this means multiple back-and-forth queries.
# 23:27 aaronpk that's kind of a hard geo search problem in general
# 23:27 tantek that sounds like something that could be summarized fairly easily
# 23:27 tommorris tantek: nope. in the city it is, but out in the countryside it isn't.
# 23:28 tommorris grew up in the countryside before moving to the big city lights. :)
# 23:28 tantek uh - public domain population maps, hash into maybe 3-4 size boxes (orders of magnitude from 2 blocks on up)
# 23:29 tommorris tantek: there are solutions, I've just been too lazy to implement them. :)
# 23:29 aaronpk as cloudmade demonstrated, it's not worth it to provide the service to small customers
# 23:30 tantek same reason as olson - name it after yourself :P
# 23:30 tommorris the current solution I'm looking at is loading OSM into Solr and using that.
# 23:31 tommorris Or the nice people who build XAPI/Overpass implement the function. ;)
verdi_ joined the channel
# 23:32 aaronpk pretty sure it's fine to use foursquare data as long as you attribute and link
# 23:33 tommorris aaronpk: personally, I like to contribute back to OSM. it might be a reasonable silo bridge for now, but ideally I'd like to get off it.
# 23:33 aaronpk My plan once I get venues and checkins on my site is to just use foursquare search for any venues that are not already local on my site
# 23:40 aaronpk running an OSM database is not something I want to do like ever
# 23:42 tantek aaronpk, ok here's the rough easy thing to do.
# 23:42 tantek for checking to see how tightly should you ask for venues (2 blocks, 1sq mile, 10sq miles etc.)
richardlitt joined the channel
# 23:44 aaronpk what projection is in tho? how do you turn lat/lng -> pixel
gr0k joined the channel
# 23:44 kylewm how many venues are you expecting to have? > 10,000?
# 23:46 aaronpk from looking at the data, i think you'd need higher-res density info for that to work well
# 23:47 aaronpk maybe tommorris has a better idea, but the data looks pretty coarse to me
# 23:48 tantek oh, just realized the ratio is ~2:1 same as latlong
# 23:50 aaronpk anyway point is even within the portland city limits you need to use drastically different search radii. cause in downtown venues are stacked on top of each other, but out in SE not as much.
# 23:50 kylewm I can see the case for searching 1 block vs. 2 blocks dependning on density, but when are you ever checking into a place a mile away?
# 23:50 tantek really? why not use 2 blocks radius as a minimum?
# 23:51 tantek this feels like the kind of problem where a few quick numerical hacks would get you something pretty good
# 23:51 aaronpk this is even a problem on foursquare. sometimes the place I'm trying to check in is way too far down the list
# 23:51 tantek I've always wondered why the checkin "screen" doesn't show a 2d view
# 23:52 aaronpk the 2d view would not be very useful. too much stuff stacked
# 23:52 tantek really? in practice 99% of things have different addresses - zero stacking
# 23:53 tantek (not apartments etc., but ground floor restaurants, theaters etc.)
# 23:53 tantek they already have that UI built in too! e.g. in explore!
# 23:53 aaronpk related, checking in from my pebble has been great cause it only takes a couple taps to find the venue in the list and click it
# 23:54 tantek not sure why the Swarm app is so much slower at getting a list of venues than Checkie is
KevinMarks and npdoty joined the channel
# 23:56 tommorris have you seen the Swarm UI inside the Apple Notifications panel?
# 23:56 tommorris it shows you the nearest thing to you and you just tap the button to check in there
# 23:57 aaronpk oh man then I can make my gps tracker do awesome things