[Mapserver-users] Tileindex questions.

[Ed McNierney]

Bob -

No, creating seamless edgematched tiles does not take care of that
aspect.  It doesn't address the issue of how MapServer determines
*which* of the 528 files needs to be used to produce a requested map.

If you're requesting an image that requires most of the source images, a
TILEINDEX won't help much.  But you'll have other problems - if you
think that's going to happen (in your case you say it won't) you should
have resampled lower-resolution images available to serve those
lower-resolution requests.  Otherwise you're going to be reading a
zillion pixels from disk only to throw away most of them.

	- Ed


Ed McNierney
President and Chief Mapmaker
TopoZone.com
-----Original Message-----
From: Bob Basques [mailto:bob.basques at ci.stpaul.mn.us] 
Sent: Friday, March 26, 2004 4:59 PM
To: Ed McNierney
Subject: Re: [Mapserver-users] Tileindex questions.

Ed McNierney wrote:

>Bob -
>
>I'm not quite sure what you mean by "composite layer" as opposed to 
>TILEINDEX, but you should certainly use a TILEINDEX.
>  
>
A composite LAYER would be a single set of data for the whole of the
LAYER coverage, vs seperate TILEs of data (500+) for the same coverage
area.

>A TILEINDEX is a very efficient way for MapServer to determine which 
>source images overlap the requested output area.
>
For discuassion purposes, I have already taken care of that aspect, the
tiles are seamless and edge matched in the generation process.

> In your example, any
>given map request *ignores* 97% of the files.  Instead of making 
>MapServer open every file to see whether it's interesting, the 
>TILEINDEX will very quickly determine WHICH 16 files you need for a 
>given request, without even touching the other 512.
>
>It is in just such cases that a TILEINDEX is very helpful.  If you were

>going to end up using all 528 images in the output anyway, there 
>wouldn't be much point in quickly figuring out which ones you needed.
>  
>
So the closer one gets to the whole coverage are for a request, the less
effective the Tile Index will be.  That's what I was looking for as a
statement.

>When a small number of tiles from a large dataset are to be retrieved, 
>a TILEINDEX is the way to go.
>  
>
Got it.  Thanks Ed.  That clears that up.

bobb