Man, I've been having bad luck with my email addresses lately... Trying again, please forgive any duplicates. --- Yeah, for 2 I would be adding the feature in the core dependency group and that would mean pretty much no matter what container anyone was using they'd end up sucking in the rpc register for the oauthpopup code. That means the gadget feature code would get sucked in too when gadgets that don't explicitly ask for it are rendered... but the gadget feature code is smaller now, so it's less of an issue.
Option 3 tries to avoid that, but is messy and is not something I'd want to set a precedent for in shindig. On Tue, Aug 21, 2012 at 7:16 PM, Stanton Sievers <ssiev...@apache.org>wrote: > Dan, can you elaborate on option 2? How would moving the oauthpopup > feature to core solve the upgrade problem? Is that simply where the > rpcregister would happen container-side? > On Aug 21, 2012 7:06 PM, "Ryan Baxter" <rbaxte...@apache.org> wrote: > > > I like option 1 but can understand why people would be upset, so option 4 > > may be your only option. Although I hope we could do option 1 post > 2.5... > > > > On Tue, Aug 21, 2012 at 11:17 AM, Dan Dumont <ddum...@us.ibm.com> wrote: > > > > > I've been looking at having the oauth popup feature make some calls > into > > > the container over rpc to handle the popup for various reasons, one of > > > which is to work around browser popup blockers. > > > The container could implement the feature as a litebox instead of a > > popup. > > > > > > This change though requires some changes that will probably break > > > unsuspecting upgraders... so my options are as follows: > > > > > > 1) Refactor oauthpopup and break unsuspecting containers when they > > > upgrade. > > > 2) Refactor oauthpopup and add it to core (it's pretty small) so that > no > > > one gets hurt on the upgrade. > > > 3) Refactor oauthpopup and add only the container part to core (this > gets > > > kinda messy... ) > > > 4) LEAVE MY OAUTHPOPUP ALONE! (mess with my own copy, but don't change > > > shindig) > > > Btw, the default implementation in my refactor calls window.open just > > like > > > the old one, only now the container is doing the window.open instead of > > > the gadget. > > > > > > What does the community think the best approach would be? > > >
