On 27 September 2011 16:33, Scott Wilson <[email protected]> wrote: > So, how best to handle this? > > 1. We can retain the broken method for generating hashes from 0.9.0 as a > "legacy mode" switchable in properties > 2. Put a big warning on the release but go ahead anyway, and promise not to > do anything so > 3. Bang head on keyboard repeatedly until a better idea comes up > > (Another option that occurred to me is to try to process each inbound widget > instance request using the old method, and if it matches, update it to use > the new one. Again we could toggle this on/off.)
The last option sounds like the best, coupled with deprecation of the old methods. On the 0.9.1 release we do 2. indicating that people have until the 0.11 or 1.0 release (whichever comes sooner) to resolve this before the deprecated behaviour is removed. During the (minimum) two release cycles this involves it is reasonable to assume that any useful data will have been migrated by admins. Note, I have a 0.8 production instance that I need to upgrade to trunk in the next few weeks. I'm happy to make this a test site. Ross -- Ross Gardler (@rgardler) Programme Leader (Open Development) OpenDirective http://opendirective.com
