Odd behaviour for random files

[Matt Mackall]

On Tue, 2011-05-17 at 20:28 +0200, Jan Olsen wrote:
> On Tue, May 17, 2011 at 20:12, Matt Mackall <mpm at selenic.com> wrote:
> 
> 
> > If I'm understanding this setup correctly (multiple users committing
> > directly to a single repository instance over a shared filesystem on a
> > WAN), then you folks probably deserve an award of some sort for "best
> > example of how to abuse Mercurial as a non-distributed system".
> >
> 
> We live in a very dynamic "internet" universe where users (ab)use
> applications/programs and extends them to the limit or even whole new usage
> methods.
> Our case might be the perfect example of how not to use Mercurial, but
> luckily the system allows us to work like that and it fits our needs
> perfectly.
> 
> We need a system to keep history of who has made which changes to what files
> and that is solved with our setup. We sit next to each other and all have
> the same intentions of keeping the systems running, so it's not like we fear
> that one of us just starts deleting random files and denies everything if
> questioned.

Oh, sure, you can make it work. And it should in fact work correctly
(modulo the reliability of your network filesystem).

But Mercurial is designed as a distributed system, so by working in this
way, you are definitely working against the grain and apparently
suffering performance-wise because of it. 

There are plenty of projects with an order of magnitude more code and
users than you and an order of magnitude less bandwidth that would laugh
heartily at "I can't commit because it's a weekday".

-- 
Mathematics is the supreme nostalgia of our time.