On Thu, Jun 13, 2013 at 8:42 AM, Dave <w...@glark.co.uk> wrote: > On 6/12/13 6:22 PM, "Joseph Gentle" <jose...@gmail.com> wrote: >> >> Regardless of where the code goes, as I've said we should redesign the OT >> system using proper TP2 types. This will enable us to build a working >> federation protocol thats better anyway. I also think we should separate out >> the OT types into a library, and make the system capable of hosting >> different kinds of data. I don't think it makes sense to prototype that >> inside the WIAB codebase. > > > Aside from an initial "toy" PoC, why not do this experimenting within the > wiab codebase? I.e. as a second pluggable OT approach. It would require > making the existing code more modular, but doing any of this work within > wiab from early on would make sure that those module boundaries and > dependencies were properly defined and implemented. > > Dave
I don't know yet where those module boundaries should be. This is an issue I came across in ShareJS - I wrote a bunch of tests early on (I did full TDD), only to discover that I wanted to change the architecture. My tests made it really difficult to iterate on how my code was assembled. I'm happy to experiment inside a directory in the WIAB repository, but I'd like to get a design review of the changes I want to make before embarking on a large refactor of the existing java code. -J
