On Wed, Mar 20, 2013 at 11:39 AM, Camillo Bruni <[email protected]>wrote:
> > 1) file based VCS, are a bit rude for Smalltalk, which operates on > smaller > > source granularity, and typicall Smalltalk based VCS handle that in more > > natural way > > not true... and just for the record, Monticello is file-based too, it does > not store any high-level information. If you store each method as separate > file as done with filetree you get exactly the same granularity as > monticello. > > see: https://github.com/pharo-project/pharo-core > > My mistake for using term "file based" loosely. I was refereeing to situation when when whole module,. package or similar larger unit of code is stored in vcs as atomic unit. As monticello can distinguish different methods and classes (probably because it knows it is storing smalltalk code and its syntax). To fit its model into the git and prevent it from making silly merges and things alike, it is necessary to store just one method into one file, model classes by directories, and stuff remaining meta info into the (currently controversial) json encoded meta files. It is quite clever and practical solution to fit Smalltalk model of work into the git storage, but I guess it is a bit unusual usage pattern for git. And certainly I would not expect anyone to do any serious browsing or looking at the code just by looking into such git directory on local disk, or through generic git web interface. My point that I was trying to make is that just plain git does not seem to be clearly superior to monticello/metacello, but combination of git and montiocello in the direction FileTree seems to be taking is. Davorin Rusevljan
