[Benchmarking] Thinking about the best effort part

[Daniel Morissette]

Andrea,

You've got a very good point here: best effort depends a lot on how much
time and disk space you put into it, and eventually we could all end up
returning pregenerated map images/tiles.

Not that I have a useful solution to share... just wanted to say that I
agree... and share the mathematical analogy that crossed my mind when I
read your email:

   lim (t -> infinity) best_effort(x) = pregenerated_images


Daniel


Andrea Aime wrote:
> Hi,
> I was looking at the test list we have and I was reasoning about the
> best effort part, in particular in terms of data transformations.
> 
> So it's pretty clear almost everybody will try to use some kind of
> pre-generalization scheme. Fine.
> I guess many will want to shave off all un-necessary attributes too
> and maybe use a different classification scheme (maybe using integer
> numbers instead of strings for the classes).
> Once you do that all line layers look pretty much the same, a line,
> a classification attribute (an eventual label).
> Since we test all the layers together, is merging those layers into
> a single one acceptable? We don't have tests against a single layer.
> Or is it considered cheating against the multilayer nature of the
> test?
> 
> And what about pre-projecting the layers into the target projection?
> Is that cheating against the reprojection nature of the tests?
> My gut says yes, but that's just me. And at the same time, isn't
> creating a pyramid a way to cheat about the multiresolution nature
> of the tests? ;-)
> 
> Cheers
> Andrea
> 


-- 
Daniel Morissette
http://www.mapgears.com/