At what point does the push model degrade?

[Bryan O'Sullivan]

Paul Sargent wrote:

> A pull model requires somebody 
> who's task is to act as gatekeeper/collator/manager of change-sets, and 
> this naturally has to be someone higher up the knowledge tree. I would 
> say that using that person for that task, and removing them from 
> development, is a significant cost.

Look at it as a way of ensuring that changes get some review from people 
who understand the system well, and it turns from a liability into a 
benefit.  If you're reviewing changes anyway, there's not really any 
added overhead, and in fact the preservation of history makes it easier 
to see what's happened than with tools that lose history.

> Again, I don't think I agree. An SVN type system doesn't require merging 
> unless the same file is changed, making that process far less likely to 
> happen.

The cost of a "hg fetch" is the same as a "svn up" whether or not there 
are conflicts.  The only difference being that if a merge occurs, you 
have to push the results of the merge back.  This is not significant.

	<b