#dev 2016-09-15

2016-09-15 UTC
#
www.svenknebel.de
edited /chat (+91) "add link to /discuss at top"
(view diff)
doesntgolf, KevinMarks_, miklb, tantek, KevinMarks and AngeloGladding joined the channel
#
cweiske.de
edited /Micropub/Servers (+48) "/* Implementation status */ known supports 1 syndication target"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+38) "/* Implementation status */"
(view diff)
KevinMarks and cweiske joined the channel
#
cweiske.de
edited /Micropub/Servers (+2) "/* Implementation status */"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+46) "/* Implementation status */"
(view diff)
cmal, loicm_, gRegorLove, KevinMarks_, KevinMarks, AngeloGladding, kants, doesntgolf and bear joined the channel
singpolyma, Zegnat and cweiske joined the channel
#
cweiske.de
edited /Micropub/Servers (+66) "/* Implementation status */ add rsvp"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+17) "/* Implementation status */"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+78) "/* Implementation status */ bookmark"
(view diff)
#
aaronparecki.com
edited /device-auth (+39) "see also"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+90) "/* Services */ mention commentpara.de"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+159) "/* Implementation status */ list commentpara.de implementation status"
(view diff)
#
cweiske.de
edited /Micropub/Servers (+27) "/* commentpara.de */ -> anoweco"
(view diff)
#
Zegnat
Surprised Kevin Marks’ rel-me verify doesn’t work in Firefox.
plindner and doesntgolf joined the channel
#
Zegnat
Are there any Firefox users at hand?
#
rascul
is a firefox user at hand
#
Zegnat
rascul, could you give https://github.com/Zegnat/verify-me/blob/master/verify-me.xpi a go for me and see if it works for you?
#
Zegnat
Some headscratching was involved, but it seems to work now. Just want to get someone else to take a look as well.
#
Zegnat
Just visit http://www.kevinmarks.com/ and see if little checkmarks pop-up next to his elsewhere links :)
#
rascul
wow kevinmarks.com takes forever for me to load
#
rascul
appears to work correctly though on rascul.xyz aaronparecki.com and bear.im
#
rascul
there we go, it seems to work correctly on kevinmarks.com
#
Zegnat
Good.
#
Zegnat
I should probably package the “buggy” version too, to see if it was a local issue or something odd with Firefox plugins
#
Loqi
I agree
gRegorLove joined the channel
#
Zegnat
rascul, if you have another minute, could you test this version too? For some reason, it does not give me any checkmarks: http://zegnat.net/tmp/verify_me-0.2.2-fx%2Ban.xpi
#
rascul
does not seem to work at all
#
rascul
oh, maybe i should enable it
#
rascul
nope, doesn't seem to be working
#
Zegnat
crazy
#
Zegnat
but thanks for confirming
#
Zegnat
Apparently Firefox does not support relative scheme URLs, because this is the entire change: https://github.com/Zegnat/verify-me/commit/acd48e6dba3cd8142fcbf11ffd0dc67a410ce9d1#diff-afb580558f47e8d8cdf35bed27f62752
#
rascul
btw this is firefox-dev 50.0a2 (2016-09-15)
#
Zegnat
I was testing on 51.0a1
#
Zegnat
Maybe I do need to get normal-Firefox for a fair test
#
rascul
umm this is aurora i guess
#
Zegnat
But apparently it will not work without scheme in either 50 or 51 then
#
Zegnat
I am on Nightly (albeit a slightly outdated one, because I only opened it now to test)
plindner joined the channel
#
KevinMarks_
That's odd
#
Zegnat
I still need to test the XPI on Firefox for Android. It might work, as I think content scripts are 100% supported there
#
KevinMarks_
How do I tell Firefox android it's OK?
#
Zegnat
I… don’t know? Is it complaining? I made sure to sign the XPI...
#
gRegorLove
Slack's new "All Unread" view is pretty nice
#
Zegnat
Alright, I can confirm the add-on works on mobile, but I could only get it installed by getting it from AMO
#
Zegnat
So either there is something wrong with my XPI on GitHub. Or Firefox for Android has some limitation I cannot find documented anywhere.
#
sknebel
Zegnat: is it signed?
#
cweiske
didn't firefox recently changed their plugin installation rules?
#
cweiske
only-from-official-store-unless-developer-build`?
cmal joined the channel
#
Zegnat
I signed it
#
sknebel
cweiske: I think you can distribute packages signed by Mozilla outside the store, but they have to go through mozilla once
#
Zegnat
Yeah, you have to provide your personal API key when you do $ web-ext sign
#
cweiske
I remember reading that - unless you're using a firefox developer build - you can't install install addons outside the mozilla store anymore
#
cweiske.de
edited /Micropub (+86) "/* Examples of Creating Objects */ link post types"
(view diff)
#
Zegnat
cweiske, any source? Self-hosted is still a thing they talk about on MDN
#
sknebel
I think 48 and later won't have a switch to turn off the signature check anymore, but I can't find anything about any version being limited to the store only
#
Zegnat
Yeah. And it is signed. I just confirmed. The download I can access on AMO gives me a file with the same md5sum as the XPI in the GitHub repo
#
Zegnat
But when I downloaded the XPI on my phone from AMO it worked, from GitHub it was blocked. So it does feel like an origin base check
#
Zegnat
I don’t really like what they are doing to browser extensions :( Firefox now requires signing by Mozilla, Safari requires you to pay for a developer certificate...
#
cweiske
.. and chrome is google
cmal joined the channel
#
sknebel
Zegnat: found a way to install it
#
Zegnat
sknebel: do tell!
#
sknebel
download it, open file:///downloads/folder, click on the .xpi there
#
Zegnat
That URI is not working for me. Do I need to discover the path to my downloa folder somewhere?
#
sknebel
yes, sorry, you have to insert your actual folder
#
Zegnat
Not doing it for me :/
#
Zegnat
file:///storage/emulated/0/Download/verify-me.xpi
#
Zegnat
But still being blocked
#
sknebel
you have to enter the folder first
#
sknebel
and then click in the folder on the file
#
sknebel
you can't enter the filepath directly
#
sknebel
because... reasons
#
Zegnat
It works
#
Zegnat
Please tell me you found an obscure article about that, sknebel, and not figured that out by trial-and-error just now :p
#
Zegnat
ping KevinMarks_ : you can now run the extension on Firefox, if you follow sknebel’s instructions
#
sknebel
Zegnat: found a related stackoverflow answer: https://stackoverflow.com/questions/36049137/installing-unsigned-xpis-in-firefox-for-android-version-42-or-newer, didn't even know firefox had directory views
#
Zegnat
It is odd that you would need to do that for a signed extension though
#
sknebel
you can whitelist domains in about:config as well
#
sknebel
haven't tried that
#
KevinMarks_
The go up a level thing worked
#
Zegnat
I will probably get the source to my extension up tomorrow - with in browser rel verification. It correctly resolves redirects now (pages linking to kevinmarks.com without www. came up verified) but there still seem to be some differences between it and indiewebify
#
Zegnat
Twitter still doesn’t resolve. The t.co URL gives a 200, not a 30* redirect.
#
Loqi
cweiske has 55 karma (5 in this channel)
#
ben_thatmustbeme
thats excellent
#
ben_thatmustbeme
actually, if I had an ability to discover server version string or something i could use that and this to auto-configure the micropub client for the person
#
ben.thatmustbe.me
edited /Micropub-brainstorming (+22) "add to micropub category"
(view diff)
#
ben_thatmustbeme
yikes, aaronpk, i think we have some template issues
#
ben_thatmustbeme
https://indieweb.org/Micropub-brainstorming the TOC is getting text running it to it for me
#
sknebel
Zegnat: are you using the fetch-api for the redirect-discovery?
#
Zegnat
I am doing a HEAD request with window.fetch(), yes
#
Zegnat
which will automatically follow redirects and turn my response.url into the final url. So then I compare that.
#
sknebel
then you can solve the twitter-issue at least in firefox by setting a different user-agent string
#
sknebel
chrome hasn't gotten around to implementing that yet, but firefox does
#
Zegnat
Wait, seriously? Twitter is giving different t.co pages depending on UA? That sounds … ridiculous
#
sknebel
fairly sure it does
#
ben.thatmustbe.me
edited /Micropub-brainstorming (+408) "/* Query for other information */"
(view diff)
#
sknebel
trying to find a reference
#
Loqi
[ben_thatmustbeme] with user agent > meta refresh, without user agent > location header
#
Zegnat
I am sticking to what I just said: ridiculous.
#
Loqi
rofl
#
Zegnat
So, what, they try to specifically cater to cURL or something?
#
sknebel
also, re-reading it seems like changing the user-agent for fetch triggers a CORS-check, so I'm not sure if it works
AngeloGladding joined the channel
#
cweiske
ben_thatmustbeme, are you building a new micropub client?
#
Zegnat
CORS should not be a problem, sknebel. My fetching happens in a background script, which means I can always fetch from any domain
#
sknebel
Zegnat: wasn't sure, that's why I mentioned it
#
KevinMarks_
Were you able to see all the urls it redirects through?
#
KevinMarks_
I think that's legitimate if it's on the redirection path
#
Zegnat
I haven’t checked, KevinMarks_. Because I get the final URL back really easy so I went with that
#
Zegnat
Aah. Alright. Sounds like I have some work left to do on the redirect resolving then ;)
#
KevinMarks_
That will help with www vs bare and other redirects
#
Zegnat
I am “borrowing” your SVGs, KevinMarks_. Can I assume those to be CC0?
#
Zegnat
(Want to ask before I go an release the source tomorrow.)
#
KevinMarks_
Yes, they're cc0 or apache or whatever yes
#
KevinMarks_
I was thinking we could add a grey tick if it links back but has no rel-me
#
cweiske
hm. known doesn't care if I send content or content[html]; it automatically detects if its html
#
Zegnat
cweiske, is that wrong? People submitting Markdown might do so as content and that can contain HTML as well. I don’t think content MUST be plain text?
#
aaronpk
that is wrong
#
Zegnat
KevinMarks_: nice, thanks for clearing up licensing :) Grey tick might be interesting
#
aaronpk
if you interpret content=<foo> as html then it's super easy to accidentally get javascript or other bad content into the post
#
aaronpk
or the non-mailcious case, it becomes impossible for the user to include characters in their post that need to be escaped
#
cweiske
e.g. when someone asks "how do I do a headline in html" and I want to reply "use <h1>"
#
cweiske
and that gets interpreted as html, but shouldn#t
#
Zegnat
Hmm. So if I am using Markdown or Wiki-Markup or something that allows for intermingled HTML, it can’t go in content= ?
#
aaronpk
well if you write your own server, you can ultimately decide to do whatever you want. but clients will be expecting they are sending you plaintext content
#
aaronpk
you *can* make that work, but you're going to have to handle the edge cases and failure modes yourself, and may lead to some unexpected results
#
aaronpk
rhiaro sometimes uses my micropub clients to edit YAML blocks so... you know
#
rhiaro
oh I'm over that phase
#
rhiaro
I was using markdown too though
#
rhiaro
Now my server expects html though
#
Zegnat
Interesting. So as far as clients should be (or are) concerned every endpoint supports only plain text (content) and HTML (content[html])?
#
voxpelli
makes sense considering that Micropub doesn't define the format but rather the mf2 parser output does
#
Zegnat
makes mental note
tantek joined the channel
#
voxpelli
transforms the html to markdown to get his markdown
#
cweiske
so user writes markdown -> client converts markdown to html -> sends to MP endpoint -> endpoint converts HTML to markdown -> saves into git -> static site generator converts markdown to html
#
cweiske.de
edited /Micropub/Servers (+46) "/* Implementation status */ known has a content bug"
(view diff)
#
aaronpk
i don't think any clients are currently "markdown-aware"
#
voxpelli
yeah, one should simply not describe the flow in a flow chart because the full flow because totally crazy
#
cweiske
aaronpk, commentpara.de will get its own "comment box" client, which will accept markdown
#
cweiske
gnight
#
voxpelli
cweiske: I btw very much do write my posts in markdown as well ;) so that flow is totally real
#
Zegnat
I am going to say goodnight as well. Have a good one, everyone!
#
Loqi
don't let the bed bugs bite
#
voxpelli
same here, good night!
#
ben_thatmustbeme
!tell cweiske yes I am building a new MP client
#
Loqi
Ok, I'll tell them that when I see them next
cmal, Loqi and KevinMarks joined the channel
#
aaronpk
? i think the unicode 8 emoji aren't supported by mysql
#
aaronpk
maybe not. but for some reason, webmention.io is failing with those.
#
aaronpk
looks fine in the chat logs tho
KevinMarks joined the channel
#
KevinMarks
Is this the same "naive utf8 always works" thing as rtl?
#
aaronparecki.com
edited /Micropub (-8) "/* h-entry */ remove slug since it's not a property of h-entry"
(view diff)
#
tantek
interesting
#
tantek
that makes me wonder, since I'm considering publishing a u-uid WITHOUT slug, and I publish a u-url WITH slug, could slug be implied ?
#
tantek
(the point of the u-uid without slug is for readers to assert uniqueness on the post even if/when I change the slug)
#
tantek
(also helps with /deleted posts - only have to keep the u-uid in the u-url, and can dump the slug completely as part of deleting all content of the post)
#
aaronpk
I don't think anybody should be parsing the path of URLs
#
tantek
this would be a simple string prefix check
#
tantek
if A is a prefix of B, then the remainder of B is the slug
#
aaronpk
that's not always the case though
#
tantek
sure, I'm trying to see if establishing a pattern would help in the future
#
aaronpk
what's the value of implying the slug based on string matching uid vs url?
#
tantek
I think the "slug is part of the id" approach will die over time, as more folks admit that the slug is or should bepart of user-editable content of the post
#
aaronpk
wordpress has a unique ID for each post, and the slug can be used in the URL
#
aaronpk
if you edit the slug, it keeps the history and redirects the old slug to the new slug
#
aaronpk
wordpress' "rel=shortlink" sometimes links to the integer ID version
KevinMarks joined the channel
#
GWG
aaronpk, did you catch up on logs?
#
aaronpk
no, i think i am going to give up on that
#
aaronpk
if there is anything important, I hope it either made it into a github issue, the wiki, or will be repeated again later
#
tantek
I plan to re-read logs from during XOXO weekend
#
tantek
since I think I missed most of that
#
GWG
aaronpk, there was a question about the license for php-comments and why it is Apache.
#
aaronpk
i definitely missed most of thursday-sunday
#
tantek.com
edited /Facebook (+1179) "/* Criticism */ add new Weaning section for how to"
(view diff)
#
aaronpk
i think that was the default license i was using at the time, so no particular reason
#
tantek
aaronpk - yes - that time period
#
GWG
aaronpk, the second question is who is looking at PRs?