I'm reflecting on Tuma MinGW future at the moment. I could push another point release for sure, but unlikely reach 1.0 milestone. Few weeks ago a developer offer me a help to turn Tuma MinGW into something like Gentoo, a self construct distro with fully automated updating/rebuild concept which I half heartedly had to reject as it will goes "by developer for developer" way. But the main reason is because I doubt its feasibility. There are many attempt with similar ideas in the past, but most went niche or failed. Me too want to go niche, making it console oriented distro (something that "nicely" oppose the Windows way).
The buildbot, I used to make buildscript to compile stuff (still for some case) but after see how diverse is build systems under Windows, I'm quite overwhelmed.
My plan is if it goes 1.0 I want to put it like msysgit repo at github, rebranding it with new name (I call it cbonk, english: tadpole) and invite collaboration. But github only offer 1GB of realistic space. Sourceforge IMO is the best at hosting big files (even ISO files) and have rsync too. I once complain at git irc about how git doing initial clone by tarballing and fetch it in plain non-resumable and inefficient http download but they had no answer for that, quite a hell for people like me. Yesterday I tried this "Github for Windows" which turn out to be a "slow motion" application based on .NET which emptied my RAM after about 100 files of 10 MB commit. I'm thinking of distribute Tuma MinGW into several repos with one as base and the rest as module. Does this need a trick?
I realize that it's bad to have repo of binaries especially at this scale maybe I should packed each application as tarball and use spkg to install/remove this way it might fit to the 1GB barrier, or maybe just get rid several compilers? I still have several to do lists to be addressed for 0.8, mostly integration work, going further I want to pimp many things using zenity to make it much friendlier without going GUIsh. hmm... clueless