#indiewebcamp 2013-01-05

2013-01-05 UTC
#
aaronpk
waves from San Francisco
#
aaronpk
tantek: yea I'm not sure what he means either
#
tantek
aaronpk - how long you in town?
#
aaronpk
leaving Sunday afternoon
#
tommorris
+1s any protocol-relative URL support.
#
tommorris
Wikipedia switched over to protocol-relative URLs a while back when they rolled out HTTPS support.
#
tommorris
very useful if you have the same resources available on both HTTP and HTTPS. something like a CSS stylesheet: <link href="//en.wikipedia.org/.../whatever.css" rel="stylesheet" />
#
tommorris
will then load using whatever the protocol the HTML document was retrieved with.
#
tantek
tommorris, what does protocol relative support mean for the relmeauth / indieauth use case however?
#
tantek
why would you want to enter //tommorris.org/ into a form instead of just tommorris.org and then let http://tommorris.org/ upnegotiate to https://tommorris.org/ if it can.
#
tommorris
no, more that if you have a link to <a href="//twitter.com/t" rel="me">Twitter</a> on tantek.com, it should be able to handle that
#
tantek
ah ok that would make sense
#
tantek
I mean, sort of?
#
tantek
Like in what case does a provider's http(s) usage depend on the indieweb site it's linked from?
#
@robodynamo
.@t Boo :/ RelMeAuth implementations (http://t.co/Wt1x72vP, https://t.co/VIZJCwUs) don't seem to accept protocol-relative URLs #indieweb
#
tantek
so for Twitter, why not always link to "https://twitter.com/t"
#
tommorris
I believe Twitter now require HTTPS
#
tantek
right
#
tantek
regardless of what site the link is on
#
tommorris
well, some people may just have an aesthetic preference for protocol-relative URLs or just like to not have to type the https: or http:
#
tommorris
or there might be a publishing platform that specifically converts things to protocol-relative
#
tantek
but isn't that bad? I mean, if you know your provider supports https, isn't it *better* if you're explicit about the https?
#
tantek
ok, theoretical publishing platform. when someone shows an example of this perhaps we can fix it (whereever makes sense)
#
tantek
peronally I think protocol relative URLs to providers is a *bad idea* and instead you should be explicit about the https when you know they support it.
#
tommorris
I think generally, Postel's law should probably apply here. ;-)
#
tantek
I don't think so. This is security related and thus the sooner an error helps you find a problem in your set-up, the better.
#
tantek
squeaky wheel and all that
#
tommorris
I'm not sure how cross-domain protocol-relative URIs work.
#
tantek
well that's the only point AFAIK
#
tantek
because if it's not cross-domain, then you just use a path-relative URI
#
tantek
starting with "/"
#
tantek
instead of "//"
#
tommorris
well, I know part of the reason we have it at wikipedia is if you login to, say, wikipedia, a cookie is set for letting you login to all the other sister sites
#
tantek
but that's different, the TLD is still the same
#
tantek
cookies don't work cross TLDs anyway
#
tommorris
well, it's more that you'd go to https://en.wikipedia.org/, it'd set a cookie for all the HTTPS wikimedia sites, and then you'd be shot off to http://en.wikinews.org and your HTTPS cookies wouldn't work.
#
tantek
I guess I'm having trouble understanding how that affects the RelMeAuth / IndieAuth flow
#
tommorris
it'd be something to check with other URI-based authentication things
#
tommorris
it'd probably be sensible to check how OpenID do it and steal that. ;-)
#
tommorris
I mean, if I put <link href="//tommorris.myopenid.com" … in the head of a page, what that ends up doing.
#
tantek
ok see this is where I'd say that's a *BUG* from a security perspective
#
tantek
and SHOULD (maybe even MUST) be rejected
#
tantek
with an error
#
tantek
indicating to the user that they should change it to:
#
tantek
<link href="https://tommorris.myopenid.com"
#
tantek
explicitly
#
tantek
You're stating a *counter* use-case (a reason for *requiring* explicit https), not a use-case.
#
tantek
(for protocol relative)
#
tommorris
so, according to OpenID2 spec, an OpenID endpoint "MUST be an absolute HTTP or HTTPS URL."
#
tommorris
I'd say there is a compelling case for requiring OpenID providers to be HTTPS only. ;-)
#
tommorris
should probably practice what he preaches and sort out HTTPS for tommorris.org ;-)
#
tantek
tommorris - indieweb HTTPS is certainly a challenge, with all the things you have to get right etc.
#
tantek
please take notes as you sort it out
tantek, danbri, brennannovak, wajiii-afk, danbri__, danbri___, friedcell, barnabywalters, josephboyle, zztr and melvster1 joined the channel
#
melvster1
hello #indieweb
#
melvster1
i just added secure encrypted chat to my homepage ... anyone interested in testing it out?