My theory on this is that you should:
- pick an implementation language;
- pick all the surrounding software (mailing list manager, revision control system, CMS for the website if necessary);
- write and release something that works (or is pretty).
Only then do you indulge in even the smallest bit of community building.
Then you can avoid the six month long startup argument about which tools to use to write the vapourware. These arguments even trump the ‘what features should we do first’ arguments!