PQM needs to take over the world

[Wichmann, Mats D]

bazaar-bounces at lists.canonical.com wrote:
> Rob,
> 
> Here's a good example of why I'd like you/us to promote PQM a bit
> more. Maybe it needs a bit more polish or doc first, but it really is
> an important part of the overall DVCS solution we're championing and
> teams are looking for. 

Maybe it's just the way we use it (incorrectly?), but for someone
used to a centralized model, the disconnected-mode behavior of
PQM will rapidly drive you to feel like PQM is your personal
enemy #1. At least it's done this for me: we operate on a 
semi-distributed model but since everything official happens off
an official tree, in particular including nightly builds that we
use to spot problems, make decisions, and drive external testing,
you have to merge stuff pretty frequently to get it into use.  
So you batch up a merge for PQM, submit it, and at some time later
you get an email, often a failure that's typically one of "star 
merge failed" (what star merge? I didn't ask for any star merge -
this one is normally a typo in paths), "branches have diverged",
or the really helpful "nothing to merge". 

Requesting a merge to a central repository is something that
MUST have immediate feedback to not drive you completely up the wall -
specifically you need to know right away when you screwed up so
you can fix it. Otherwise, you're mentally off on something 
completely different assuming everything's okay and have to 
interrupt that to go figure out what to fix. The concept of a 
patch "queue" that's processed periodically is a real pain point.
A cvs/svn afficionado will not be happy with it...