[Benchmarking] Vector layers: towards a definitive list of styles and layers

[Daniel Morissette]

Andrea Aime wrote:
> 
> In the last meeting we decided that each layer should be tested
> against shapefiles, postgis, sde over postgis (right?)
> This makes 3 tests per layer, over 4 servers (with ESRI not able
> to connect to postgis without SDE intermediation).
> 

This is also my understanding.

> We also decided we'll run some tests against PNG24 with full 
> antialiasing, and some others against PNG8 (or was it GIF)
> without antialiasing.
> 

I believe we decided to go with GIF instead of PNG8 in the end. From the 
IRC logs:

(10:45:20 AM) danmo: edited proposal: "be jpeg for imagery, gif for 
non-antialised vector provided we do any, and png24 for antialiased"
(10:45:28 AM) jmckenna: k
(10:45:35 AM) danmo: does everyone support all three?
(10:45:55 AM) aaime: GS and MS do afaik
(10:46:03 AM) augusttown_: ESRI supports all three
(10:46:03 AM) danmo: ESRI, ERDAS?
(10:46:08 AM) sowston: ERDAS, no problem
(10:46:15 AM) danmo: excellent, so we have a deal!
(10:46:16 AM) jmckenna: sweet


> In the first meeting we also observed how making many tests
> against the road layers was considered not very productive.
> 
> So here is a proposal of what we could test:
> 
> - texas_roads_styled, png24, full antialiasing, no labels,
>   with the style suggested here:
>   http://wiki.osgeo.org/wiki/Texas_roads_styled
> - point_layer_styled using gnis_pop, png24, full antialiasing,
>   style tbd (see other mails)
> - polygon_layer_styled, png8 (gif), no antialiasing, a mix
>   or solid fill and textured fills maybe?

We could use that as a starting point. If we have time left at the end 
(which I doubt) we could always add more styling cases.

I think it could be interesting to test roads with labels as well (do 
all servers support curved labels?), but as I just wrote we can keep 
that for the end if we have time.

> - do we want to run a multilayer test with all the above
>   in a single request? Same styling, png24, full antialiasing
> 

Why not. That doesn't involve any additional data/style prep work, just 
another set of requests to run with all layers turned on. I don't expect 
any surprises here, but we never know.

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