[darcs-users] Benchmarking "get"

[Max Battcher]

Ian Lynagh wrote:
> Hi all,
> 
> I've done some benchmarking of "get", comparing camp, git, darcs1 and
> darcs2. Although my goal was to see how camp compared, you may be
> interested in how darcs compares in various scenarios:
> 
> http://projects.haskell.org/camp/benchmarking/get.shtml

It's certainly interesting to see that regardless of Git's weird inode 
manipulation madness (which seems dangerous and prone to cross-platform 
misfortune to me) for local gets resulting in "blazing" speeds that it 
is indeed slower in HTTP gets than darcs 2 with hashed/darcs-2 repos, 
which was my own surmise given anecdotal experience.  I didn't think to 
try a git fast-import of the same repo to do a similar<->similar repo 
experiment as you have. Thanks.

I seriously believe that most git users are hugely underestimating the 
speeds of their git operations...  and I'm slowly starting to worry if 
even some of us that are day-to-day darcs users are getting stuck in the 
"darcs is slow" mentality and continually overestimating the speeds of 
our darcs operations. Not that I'm saying that performance isn't a 
problem or couldn't improve, just that we've got a potential for getting 
lost in the very "our performance will always suck" mentality that git 
nuts want us to be in, and that anti-Haskell people have always claimed.

Particularly, because we are mostly benchmarking darcs against the 
various versions of itself and it's easy to lose sight of the (lack of) 
speeds of the competition (and easy for outsiders to conflate issues 
like darcs-1 repos are slow in darcs >= 2.0 with "*my* darcs is always 
slow").

I don't have any ideas right now on things to do/focus on, but if darcs 
were to pick up a Marketing Manager at some point, I expect a big task 
for her would be to sort through what I just wrote and come up with some 
sort of plan.

--
--Max Battcher--
http://worldmaker.net