You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@shindig.apache.org by Brian Eaton <be...@google.com> on 2008/08/14 08:31:53 UTC

next gen OAuth in Shindig

On Mon, Aug 11, 2008 at 12:50 PM, Kevin Brown <et...@google.com> wrote:
> Not necessarily. As we learned with REST, validating the spec within the
> reference implementation is a very useful exercise to go through before
> voting in a standard.
>
> I'd rather have shindig eat the pain of upgrading than live with a crappy
> spec.

Funny you should mention that trade-off.

I've just sent a proposal to the spec mailing list:
http://groups.google.com/group/opensocial-and-gadgets-spec/msg/4f65d43e0446abbc.
 I don't think it would be a good idea to make the proposal a part of
the opensocial spec until we've got a working implementation,
preferably backed by one or more social containers saying "wow, that's
compelling, we're investing time to make it work for us."

Implementing this in Shindig would mean replacing most of the signed
fetch code with the OAuth equivalent (though of course in a backward
compatible fashion.)  That's potentially disruptive, but it has some
compelling advantages.  In particular, it leads directly to support
for HMAC signatures in signed fetch.  Non-Shindig containers have that
already, so having it in Shindig would be nice.

To summarize the pros:
- we get HMAC sigs in signed fetch
- we reduce code duplication
- we probably get social containers to deploy full OAuth

To summarize the cons:
- anyone deploying Shindig w/ signed fetch needs to sync up with the
changes.  I don't think this will be too painful, but it will mean
keeping up with changes as they happen.

Thoughts on this?

Cheers,
Brian

Re: next gen OAuth in Shindig

Posted by Gonzalo Aune <go...@gmail.com>.
Worth the effort, right? =), i think its a good way to look forward to the
new version =)

G.-

On Thu, Aug 14, 2008 at 4:44 AM, Kevin Brown <et...@google.com> wrote:

> If existing (RSA) signed fetch continues to work, it's probably fine.
> Breaking that would break many social gadgets.
>
> On Wed, Aug 13, 2008 at 11:31 PM, Brian Eaton <be...@google.com> wrote:
>
> > On Mon, Aug 11, 2008 at 12:50 PM, Kevin Brown <et...@google.com> wrote:
> > > Not necessarily. As we learned with REST, validating the spec within
> the
> > > reference implementation is a very useful exercise to go through before
> > > voting in a standard.
> > >
> > > I'd rather have shindig eat the pain of upgrading than live with a
> crappy
> > > spec.
> >
> > Funny you should mention that trade-off.
> >
> > I've just sent a proposal to the spec mailing list:
> >
> >
> http://groups.google.com/group/opensocial-and-gadgets-spec/msg/4f65d43e0446abbc
> > .
> >  I don't think it would be a good idea to make the proposal a part of
> > the opensocial spec until we've got a working implementation,
> > preferably backed by one or more social containers saying "wow, that's
> > compelling, we're investing time to make it work for us."
> >
> > Implementing this in Shindig would mean replacing most of the signed
> > fetch code with the OAuth equivalent (though of course in a backward
> > compatible fashion.)  That's potentially disruptive, but it has some
> > compelling advantages.  In particular, it leads directly to support
> > for HMAC signatures in signed fetch.  Non-Shindig containers have that
> > already, so having it in Shindig would be nice.
> >
> > To summarize the pros:
> > - we get HMAC sigs in signed fetch
> > - we reduce code duplication
> > - we probably get social containers to deploy full OAuth
> >
> > To summarize the cons:
> > - anyone deploying Shindig w/ signed fetch needs to sync up with the
> > changes.  I don't think this will be too painful, but it will mean
> > keeping up with changes as they happen.
> >
> > Thoughts on this?
> >
> > Cheers,
> > Brian
> >
>

Re: next gen OAuth in Shindig

Posted by Brian Eaton <be...@google.com>.
Agreed.

On Thu, Aug 14, 2008 at 12:44 AM, Kevin Brown <et...@google.com> wrote:
> If existing (RSA) signed fetch continues to work, it's probably fine.
> Breaking that would break many social gadgets.
>
> On Wed, Aug 13, 2008 at 11:31 PM, Brian Eaton <be...@google.com> wrote:
>
>> On Mon, Aug 11, 2008 at 12:50 PM, Kevin Brown <et...@google.com> wrote:
>> > Not necessarily. As we learned with REST, validating the spec within the
>> > reference implementation is a very useful exercise to go through before
>> > voting in a standard.
>> >
>> > I'd rather have shindig eat the pain of upgrading than live with a crappy
>> > spec.
>>
>> Funny you should mention that trade-off.
>>
>> I've just sent a proposal to the spec mailing list:
>>
>> http://groups.google.com/group/opensocial-and-gadgets-spec/msg/4f65d43e0446abbc
>> .
>>  I don't think it would be a good idea to make the proposal a part of
>> the opensocial spec until we've got a working implementation,
>> preferably backed by one or more social containers saying "wow, that's
>> compelling, we're investing time to make it work for us."
>>
>> Implementing this in Shindig would mean replacing most of the signed
>> fetch code with the OAuth equivalent (though of course in a backward
>> compatible fashion.)  That's potentially disruptive, but it has some
>> compelling advantages.  In particular, it leads directly to support
>> for HMAC signatures in signed fetch.  Non-Shindig containers have that
>> already, so having it in Shindig would be nice.
>>
>> To summarize the pros:
>> - we get HMAC sigs in signed fetch
>> - we reduce code duplication
>> - we probably get social containers to deploy full OAuth
>>
>> To summarize the cons:
>> - anyone deploying Shindig w/ signed fetch needs to sync up with the
>> changes.  I don't think this will be too painful, but it will mean
>> keeping up with changes as they happen.
>>
>> Thoughts on this?
>>
>> Cheers,
>> Brian
>>
>

Re: next gen OAuth in Shindig

Posted by Kevin Brown <et...@google.com>.
If existing (RSA) signed fetch continues to work, it's probably fine.
Breaking that would break many social gadgets.

On Wed, Aug 13, 2008 at 11:31 PM, Brian Eaton <be...@google.com> wrote:

> On Mon, Aug 11, 2008 at 12:50 PM, Kevin Brown <et...@google.com> wrote:
> > Not necessarily. As we learned with REST, validating the spec within the
> > reference implementation is a very useful exercise to go through before
> > voting in a standard.
> >
> > I'd rather have shindig eat the pain of upgrading than live with a crappy
> > spec.
>
> Funny you should mention that trade-off.
>
> I've just sent a proposal to the spec mailing list:
>
> http://groups.google.com/group/opensocial-and-gadgets-spec/msg/4f65d43e0446abbc
> .
>  I don't think it would be a good idea to make the proposal a part of
> the opensocial spec until we've got a working implementation,
> preferably backed by one or more social containers saying "wow, that's
> compelling, we're investing time to make it work for us."
>
> Implementing this in Shindig would mean replacing most of the signed
> fetch code with the OAuth equivalent (though of course in a backward
> compatible fashion.)  That's potentially disruptive, but it has some
> compelling advantages.  In particular, it leads directly to support
> for HMAC signatures in signed fetch.  Non-Shindig containers have that
> already, so having it in Shindig would be nice.
>
> To summarize the pros:
> - we get HMAC sigs in signed fetch
> - we reduce code duplication
> - we probably get social containers to deploy full OAuth
>
> To summarize the cons:
> - anyone deploying Shindig w/ signed fetch needs to sync up with the
> changes.  I don't think this will be too painful, but it will mean
> keeping up with changes as they happen.
>
> Thoughts on this?
>
> Cheers,
> Brian
>