I would like the darcs team to host a performance hacking contest at some point.

The rough idea is that whoever makes the biggest dent in our Standard Darcs
Benchmarks (and gets their patches accepted into darcs) gets a small prize and
bragging rights.

The purpose of this contest is to increase participation in darcs hacking from
the wider Haskell community.  To do this, we give them something very concrete
and well-defined that they can set their sights on.

The prize and the scale of the contest should likely be small to begin with,
just so that the organisers can gain some experience in running this sort of thing.

I have checked with Don who has checked with O'Reilly who in turn have suggested
that they *may* be willing to contribute a copy of Real World Haskell as a prize.

Action: when we get more details on how, when, etc to run this thing, we should
formulate a more concrete request, which we can then pass on to O'Reilly via Don

Other prizes: the usual darcs t-shirt and mug, maybe something from AllMyData.org

Timing: perhaps after the 2009-01 release and before the 2009-03 darcs hacking
sprint?

Some vague open questions:
1. How should the contest work? 
2. How long should the contest run?
3. Who should run this thing?  
4. How many entries are we realistically prepared to field?
5. What infrastructure do we need to set up?
6. When should we do this?

How it works
------------
We give people a snapshot repo that they can darcs get, and to which they can
send darcs patches.  Darcs patches can go a special email address so that we
know they are contest-related.  We give them something like N time (say 1 week)
to get to grips with the darcs code, hack and submit patches.

The organisers collect all patches, passing them up to the maintainers as we go
along.  We can do the reviewing in secret if need be :-) just the maintainers
talking to the coders; or we could do it in the open to force people to look at
different things.

When the deadline has passed, and all contest patches are reviewed, the
organisers do competitive benchmarking on one bundle per participants (normally,
this would consist of all contest patches sent by the participant, but maybe we
can let them give us special instructions).

How long it should run
----------------------
We don't want this to drag on.  I think maybe a week should be fine (spanning
two weekends).  Also we want to send the message that this is easy.

Hopefully we will converge towards a nicely fleshed out plan, and we are sure we
can do this thing.

bugs:
> 
> New submission from Eric Kow <kowey@darcs.net>:
> 
> I would like the darcs team to host a performance hacking contest at some point.
> 
> The rough idea is that whoever makes the biggest dent in our Standard Darcs
> Benchmarks (and gets their patches accepted into darcs) gets a small prize and
> bragging rights.
> 
> The purpose of this contest is to increase participation in darcs hacking from
> the wider Haskell community.  To do this, we give them something very concrete
> and well-defined that they can set their sights on.
> 
> The prize and the scale of the contest should likely be small to begin with,
> just so that the organisers can gain some experience in running this sort of thing.
> 
> I have checked with Don who has checked with O'Reilly who in turn have suggested
> that they *may* be willing to contribute a copy of Real World Haskell as a prize.

Further info:

   "we'll almost always throw in a free book for a prize or
    something like that."

   "I'm happy to send books to give away as prizes or for review. Just let
    me know where to ship it along with a phone number."

On Mon, Nov 3, 2008 at 12:10 PM, Don Stewart <dons@galois.com> wrote:
> bugs:

>> I have checked with Don who has checked with O'Reilly who in turn have suggested
>> that they *may* be willing to contribute a copy of Real World Haskell as a prize.
>
> Further info:
>
>   "we'll almost always throw in a free book for a prize or
>    something like that."
>
>   "I'm happy to send books to give away as prizes or for review. Just let
>    me know where to ship it along with a phone number."
>

Could we make it a *signed* copy?  That would make it more personal
and cool, IMO.

Also, I'd be willing to donate money to cover or help cover the cost
of a prize.  I'd be quite happy to put up $50.

Jason

allmydata.org will totally make a custom coffee mug with a thank-you note on it
for this contest. :-)

dagit:
> On Mon, Nov 3, 2008 at 12:10 PM, Don Stewart <dons@galois.com> wrote:
> > bugs:
> 
> >> I have checked with Don who has checked with O'Reilly who in turn have suggested
> >> that they *may* be willing to contribute a copy of Real World Haskell as a prize.
> >
> > Further info:
> >
> >   "we'll almost always throw in a free book for a prize or
> >    something like that."
> >
> >   "I'm happy to send books to give away as prizes or for review. Just let
> >    me know where to ship it along with a phone number."
> >
> 
> Could we make it a *signed* copy?  That would make it more personal
> and cool, IMO.
> 

Certainly. I'm not sure how we'd get all the authors to sign it though.

-- Don

I think we should talk about this at one of the sprints once we've merged in the
hashed-storage work.

Deferring further.