[Shootout-list] preserving historical data

Bengt Kleberg bengt.kleberg@ericsson.com
Wed, 29 Sep 2004 14:14:37 +0200


Brandon J. Van Every wrote:
...deleted
> Some HW+OS platforms are sufficiently similar to make comparisons.  For
> instance, I feel comfortable comparing the performance of any 2 Intel
> boxes on Windows 2000 and Windows XP.  For a lot of kinds of tests, it's
> not going to be that different.

i am sure you mean something different with this than what i am reading 
into it.

let me check that i have misunderstood by asking you to confirm what i 
mistakenly belive you to have written:
would you feel comfortable comparing the performance of any 2 Intel 
boxes, no matter their hardware (apart from the intel bit), as long as 
they where running Windows 2000 and Windows XP?


>>>Who said g++ is no longer supported?  Yet it has changed
>>
>>over the years.
>>
>>why do you care about the realtive performance of g++ today
>>versus g++ of yesterday?
> 
> 
> Because someone's using an older g++ for something somewhere.  Also
> people argue about whether such-and-such compiler optimization does any
> good, and I think historical data on such things is relevant to the
> argument.

ok, here is where we think differently. i would never expect an old test 
to be comparable to a new one. i would want the old compiler to be run 
side-by-side with the new, to be able to take any interest.
i think your way is just speculation. even side-by-side is mostly 
smoke-and-mirrors. imho.


...deleted
> just don't see why you think old data should be thrown away.  It's a lot
> of work to acquire such data, and it isn't expensive to archive.  If the

since it is a computer that is generating the data, i fail to see how it 
is hard work to generate it. and i am lazy.


> I think you underestimate the degree to which people form impressions
> about things based on their experiences of several years ago, rather
> than what's true today.  It's better to be able to say, "No, look, that
> was true 4 years ago when you last looked at it.  Here's the data."

in this case i think that an archive would make it more likely for 
everybody concerned to ignore what is true today.

i am aware of how much people form their impressions about things based 
on their experiences of several years ago. i still prefer netgear to 
d-link because of what happened 4 years ago. i know it is stupid, yet i 
can not stop.


bengt