#dev 2017-10-05

2017-10-05 UTC
[cleverdevil], snarfed, tantek, leg, jjuran, snarfed1, snarfed2, eli_oat and dougbeal|iOS joined the channel
#
martymcgui.re uploaded /Special:Log/upload "uploaded "[[File:hwc-baltimore-2017-10-04.jpg]]" https://indieweb.org/File:hwc-baltimore-2017-10-04.jpg"
renem, Kaja_, tbbrown, [michellejl], jjuran, jeremycherfas and j12t joined the channel
#
Zegnat !tell aaronpk Mmmmmmm https://webmention.io/aaronpk/webmention/rZJ-x5gdLSbhcqoSnH3u
#
Loqi Ok, I'll tell them that when I see them next
#
Zegnat Oh, n/m aaronpk. This is the https://github.com/w3c/webmention/issues/91 issue where microformats accept data.u-in-reply-to but Webmention doesn’t.
#
Loqi [Zegnat] #91 “other similar links” is a vague description for what constitutes getting mentioned in HTML
[kevinmarks] joined the channel
#
[kevinmarks] blockchain << https://medium.com/quinn-norton/a-ledger-and-a-network-89d7809e7416
#
Loqi ok, I added "https://medium.com/quinn-norton/a-ledger-and-a-network-89d7809e7416" to the "See Also" section of /blockchain
#
loqi.me edited /blockchain (+70) "[kevinmarks] added "https://medium.com/quinn-norton/a-ledger-and-a-network-89d7809e7416" to "See Also"" (view diff)
j12t, wagle and [markmhendrickso joined the channel
#
petermolnar so, I've started playing around with full text search in a bit more depth, and the summary is: language support is utterly terrible in most of the simply accessible fts libraries, however, sqlite does a usable and extremely simple job
#
petermolnar stemming and tokenizing Hungarian text seems to be a good test to see if a library is usable at all
#
Zegnat Looking forward to reading the blogpost if you end up implementing it petermolnar.
#
petermolnar doesn't worth a blog post
#
petermolnar it's about 50 lines of code
#
petermolnar or a bit more
#
petermolnar the problem with Hungarian (similarly, Finnish, and a good couple of other languages) is that the root word sometimes changes depending on the "target" of the phrase; eg: roof = tető; on the roof = tetőn; it's roof = teteje; on it's roof = tetején;
#
petermolnar so when you search for "tető" you should find all of those things
#
petermolnar in theory, of course
#
petermolnar which is more naturaly language search than simple full text search
#
petermolnar and is quite far from reality right now
#
petermolnar oh, I suddenly feel in 2003, I ended up in sqlite mailing list archive trying to find answers :)
#
petermolnar this looks promising: https://github.com/abiliojr/fts5-snowball
#
Loqi [abiliojr] fts5-snowball: Snowball tokenizer for SQLite FTS5
jeremych_ and [kevinmarks] joined the channel
#
[kevinmarks] I remember writing code to look for stems a while back. It's weirdly a bit like writing a compressor
#
[kevinmarks] If you do a lzw type compression on text and look at the tables it makes you get a set of syllable candidates
eli_oat, j12t, singpolyma and snarfed joined the channel
#
aaronpk Zegnat: what do you mean "microformats accept data.u-in-reply-to but Webmention doesn’t"
#
Loqi aaronpk: Zegnat left you a message 6 hours, 5 minutes ago: Mmmmmmm https://webmention.io/aaronpk/webmention/rZJ-x5gdLSbhcqoSnH3u
#
Zegnat data.u-in-reply-to gets parsed by mf2 parsers so my h-entry is a reply to your post. But a Webmention endpoint doesn’t accept a link within data[value] as a proper link so my h-entry doesn’t link to your post.
#
Zegnat webmention receiver discards the mention, it doesn’t show up as a comment on your end, even though the mf2 on my h-entry is correct
John___ joined the channel
#
aaronpk IIRC microformats parsers don't resolve URLs in <data> so you really shouldn't be using that for links
#
aaronpk http://pin13.net/mf2/?id=20171005152359398
#
Zegnat https://github.com/w3c/webmention/issues/91 and the therein linked issues on Telegraph and mention client detail it
#
Loqi [Zegnat] #91 “other similar links” is a vague description for what constitutes getting mentioned in HTML
#
aaronpk is there a reason you don't visibly link to the thing you're replying to in your post?
#
Zegnat mf2 parser specifically allows data[value] for u- properties. https://github.com/microformats/microformats2-parsing/issues/10 will make it so URLs will be resolved too.
#
Loqi [Zegnat] #10 u- parsing should always do relative URL resolution
#
aaronpk oh yeah i remember that thread
#
Zegnat I used to do stuff like “Aaron writes in his blogpost about X …” to preface my comment, and add the link in there. But that would also make that show up in people’s comment displays and looked really silly.
#
aaronpk why not a reply context, outside the content of your post
#
sknebel e.g. see my bars on top of posts for minimal variant
#
Zegnat I will do that on the next iteration of the blog design, yes.
#
Zegnat I currently author my posts in HTML, and that HTML gets assigned to the e-content property. So I can’t author anything outside of content.
#
Zegnat Technical issue.
#
aaronpk just has to be outside the e-content
jonnybarnes joined the channel
#
Zegnat Yep, which would require a new field in my editor and back-end storage system. The latter requiring a rewrite.
#
Zegnat So it isn’t coming until the next iteration of the blog.
#
aaronpk wow k
#
sknebel oh, you don't have any frontmatter at all?
#
Zegnat Nope
#
sknebel hm
#
sknebel I guess than invisible <a> tags are your best bet
#
Zegnat I don’t like doing hidden links though, as you have to make sure they are outside of the tab context and stuff to support non-graphical users.
#
aaronpk so this is the code in XRay that is used to check whether the link exists for an incoming webmention
#
aaronpk https://github.com/aaronpk/XRay/blob/master/lib/XRay/Formats/HTML.php#L44
#
aaronpk i'm specifically looking at <a> <img> <video> and <audio> tags
#
aaronpk this is before it's handed off to the microformats parser
#
aaronpk i suppose i should expand that to include more tags?
#
Zegnat You follow the current recommended tags from the Webmention spec, which is fine.
#
Zegnat I just have to deal with changing my markup to match the Webmention spec if I want to send mentions, I guess.
#
Zegnat But I opened an issue against the Webmention spec because I personally find it very vague what elements should be considered a “link”
#
sknebel list of tags that have url-typed properties would be interesting to check
#
aaronpk i would assume HTML defines that somewhere
#
Zegnat One omission that directly springs to mind is `area` for instance.
#
Zegnat “4.8 Links” in the HTML spec is a, area, and link
#
Zegnat It’s all in the issue ;)
#
Zegnat sknebel, is there such a “url” category in the spec? I couldn’t find any when I was writing the issue :/
#
sknebel Zegnat: not sure, that's why I asked
#
aaronpk https://stackoverflow.com/questions/2725156/complete-list-of-html-tag-attributes-which-have-a-url-value
#
aaronpk which links to https://www.w3.org/TR/html5/index.html#attributes-1
#
Zegnat *[href], *[src], object[data] - was the best list I could think of.
#
Zegnat Oh, we might want to add *[cite] now that I see it on that attributes list
#
Zegnat Hmm, looks like there is a bit more that allows for URLs. Ugh. (Note that this will not fix my issue with the data element, but again, that’s on me.)
[kevinmarks] joined the channel
#
[kevinmarks] Well area would make sense for person tagging an image
#
aaronpk so i guess to bring this back to microformats, should microformats be parsing URLs from attributes that don't allow URLs?
#
sknebel they'll always do that at some point, unless you remove the rule that the text content is parsed at the end
#
aaronpk yeah i guess cause everything ends up as a string in the end there's no real difference
#
sknebel yeah. we have a bunch of url-related issues right now: https://github.com/microformats/microformats2-parsing/issues/9 https://github.com/microformats/microformats2-parsing/issues/10 https://github.com/microformats/microformats2-parsing/issues/7
#
Loqi [sknebel] #9 "return the normalized absolute URL" for invalid URLs?
jeremycherfas joined the channel
#
aaronpk in other news, i really need to restore the facebook og tags on my posts so they look better on twitter/facebook
#
Zegnat I don’t mind microformats and Webmentions not being 100% compatible, I’ll adapt in the next blog iteration anyway. That’s a minor thing to keep in mind.
#
Zegnat As long as the Webmention spec is clear in what receivers are going to understand, so we can work against that
#
sknebel yeah, not sure how much of that fits in the spec vs is external practice. at least some applications will have different needs
#
sknebel and e.g. the spec describes how to handle JSON, which to my knowledge nobody has implemented
#
sknebel (It's a fair question why it's in the spec then, but...)
j12t and amz3` joined the channel
#
bear hey all FYI Git CVE out -- https://people.canonical.com/~ubuntu-security/cve/2017/CVE-2017-14867.html
#
Zegnat The receiver should send an HTTP accept header, sknebel. So if one accepts json you could send that back. Maybe even send the parsed mf2? Though nobody does it.
#
sknebel Zegnat: of course you could, but that's my point, despite the spec describing it nothing does it
#
Zegnat That part of the spec always feels a little normative to me. But with HTML, elements are even included in the implementation report, so that sounds a whole lot more definitive.
#
sknebel fair enough
#
Zegnat For a normative part, to point out the spec is technology agnostic, I think its fine.
#
aaronpk this is where JSON was added https://github.com/w3c/webmention/issues/5 https://indieweb.org/wiki/index.php?title=Webmention&diff=23671&oldid=23666
#
Loqi [azaroth42] #5 Consider expanding verification step 2
#
aaronpk here's the discussion from the face-to-face meeting https://www.w3.org/wiki/Socialwg/2015-12-02-minutes#discuss_Webmention_to_take_it_to_First_Public_Working_Draft
#
Zegnat Thanks for the background aaronpk, I’ll have to have a read :)
j12t, Kaja, KartikPrabhu, tbbrown and gRegorLove joined the channel
#
martymcgui.re edited /User:Martymcgui.re (+72) "/* archives by month/year and other discoverability */ note about navigation" (view diff)
[grantcodes] joined the channel
#
[grantcodes] Not sure if anyone is using it apart from myself and keithjgrant but I have a new version of my micropub js library up with greatly improved error handling: https://www.npmjs.com/package/micropub-helper
[pfefferle], Kaja, [miklb], tbbrown and KartikPrabhu joined the channel