Lynn W. Taylor wrote:
> The benchmark affects the estimated run time, and the amount of work 
> downloaded.  It affects credit, and credit is "fun" but it's not science.

However, the people that care about the credits consider them to be a 
lot more than just a bit of frivolous "fun". Have you seen the volumes 
of discussion about credits?!...

Also, the various stats taken and presented for all manner of aspects 
for the credits awarded very well /could/ be a science *IF* those 
credits *DID* actually represent anything tangible or accurate.

How many stats sites are there now for Boinc?...

At the moment, I suspect that the error bars might mean a little for s...@h 
(and any similar FLOPs intensive project counted in the same way). 
However, for other projects, I expect the error bars would drive away 
most of the competitive crunchers.

And as for the argument that not many people would leave should the 
credits be abandoned, it would be 'interesting' to see that tried and to 
see what proportion of the results the competitive crunchers do indeed 
provide, for free. Try a zero credits month as an experiment?


> As you correctly said, it will measure something about the quality of 
> the network.

And the credits *could* be used as a *useful scientific measure* of the 
success of Boinc, the efficiency of certain methods over other methods, 
the effectiveness of some hardware over other hardware, and who knows 
what else, *if* we had some *accuracy* to what is being added up. 
Calibration? Consistency?

At the moment, all that the crude benchmark is useful for is to say that 
Boinc can be considered to be a "pseudo-supercomputer" of a potential 
"xxx Tera-whetstones". Good for grabbing headlines and I guess good for 
impressing ignorant bean-counters bureaucrats (and an unfortunate 
distracting necessity).


> ... (and you apparently didn't read) is that the science 
> application produces a result, and that result is either valid (the... 

Can we all keep off the ad-hominem please and keep this "scientific"?

Aside: How do you rate the Monte-Carlo type simulations?... Or indeed 
such as the CPDN simulations that can sometimes be sensitive to which 
hardware is running the simulation? How 'correct' is 'correct'?...

Do we "reward" for the result returned, or for the resource that a 
project has usefully utilised?


If the credits are such a bit of "fun" and "fluff", should we avoid all 
the wasted discussion and abandon them now?...

Regards,
Martin

-- 
--------------------
Martin Lomas
m_boincdev ml1 co uk.ddSPAM.dd
--------------------
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to