Matt,
Thanks for weighing in and hopefully taming this snarl. As the person
who might have posed the question originally, I figured I at least
owed a bit of constructive critique.
What can we change about OAuth that would make this better?
1) User experience - it's been echoed a number of times in this board,
so i won't beat the dead horse... much... but basic auth *is*
much simpler for the user.
The reason is obvious: oauth requires a browser. In many platforms
(especially mobile) that's a painful burden.
The PIN code seems to be ignoring the elephant in the room. It solves
some problems, but at a further cost to the user experience, giving an
even larger advantage to existing basic-auth apps.
You've made the pill even more effective, but so bitter that your
patients can't swallow it.
Providing a scheme that does not require a browser is an obvious way
to cut this gordian knot.
2) Application authentication - if your goal is to *identify* each
open source application in a *trusted* way, then I think you could be
in for an uphill battle. There are obvious technical challenges,
however the larger problem is that in OSS there is no way to define
"each application." There will be many binaries for different
platforms and people will fork it on github just because they can.
You cannot identify each of these variants as the same when they could
be different.
And that places a burden on the user experience. When a user grants
access to "application x" -- what does that mean exactly? Just that
binary? Just this release? Only from a specific trusted company?
How do you explain to the user where this subtle line is drawn in a
box he'll click through in less than a second?
I personally don't see an obvious solution to this problem. It seems
to be a UI challenge and a technical challenge. In cases like that it
seems prudent to question your goals and check feasibility.
Isaiah
YourHead Software
supp...@yourhead.com
http://www.yourhead.com
On Jul 1, 2009, at 9:46 AM, Matt Sanford wrote:
Hello again,
I do not recommend having individual end users register for
consumer keys/secrets [1] under any circumstances. So, with that out
of the way, let us focus the discussion a bit more. What can we
change about OAuth that would make this better? A complete technical
[2][3] discussion on what we could add that would make this better
is welcomed. More than welcome, it's pretty much required before we
can help.
The PIN flow was the first addition to address the inherent
insecurity of the consumer key/secret all desktop applications [3].
This stopped applications from being able to collect tokens by using
the consumer key/secret and a confidence scam (phishing like
"GoodApp needs you to re-approve us"). It sounds like there is a
fervent need for something more … what do people suggest? We're
working hard on the problem but many of you are working from the
consumer standpoint and probably have great feedback.
Please, take your time and write a well thought out reply. One-
line snarky comments, while fun to write and sometimes to read,
steal time from everyone reading the list, including all of the
Twitter API engineers. They also make the list look less inviting to
new comers.
Thanks;
– Matt Sanford / @mzsanford
Twitter Dev
[1] - People installing an instance of your server-side app are not
'end users', but other developers
[2] - Not open-source hand waving.
[3] - Closed source desktop apps have the same problem. Reverse
engineering is not stopped when you don't include the source.
On Jul 1, 2009, at 9:33 AM, DWRoelands wrote:
Actually, since Twitter has said that Basic Auth will eventually go
away, OAuth is going to be the only choice for authentication.
Twitter has forced the choice by implementing OAuth in the way that
they did.
Why should a user who chooses to support open source by using an
open-
source Twitter client be punished by having to go through extra hoops
that users of closed-source clients don't have to endure?
Forcing users of open source Twitter clients to register their
individual installations as Twitter applications is not a viable
solution. Matt Sanford has even said so.
No one is asking for "easy". I just want open source Twitter desktop
clients to be able to compete with closed-source versions when it
comes to security. Right now, that's not possible because of
Twitter's implementation of OAuth.
Regards,
Duane
On Jul 1, 11:23 am, Andrew Badera <and...@badera.us> wrote:
But that's the choice you're forced to make by OAuth, not Twitter.
And
it is YOUR choice. Personally, I would probably use the conventional
mechanisms of open source: mailing lists, special interest and user
groups. Pound the pavement and promote yourself. Who said it was
going
to be "easy"?