I use precise (x,y,z) rather than tile coords for all my objects (some walls included). I update the tile collision map depending on the position and size of each object. The "size" is the size of what you call "the anchor" (Why do I feel like Dr.Evil with all those quotes ).. I.e. if a tile is 40 world units, and a tree base is 80x80, it takes up 2x2 tiles. This is not 100% accurate since a tree with base 60x60 would also take up 2x2 tiles, but that I can live with.

Reg. drawing order, simply transform world(x,y,z) to "screen" (x,y,z) using isometric mapping, and sort according to screen Z. Obviously this is in effect midpoint sorting, so it won't work for cyclic overlap.

Reg. not splitting your objects; Unless you want to use a Z-buffer, you have to split your object into "convex volumes" to avoid problems with cyclic overlap.

/Niels

Sieggy: Something I thought you might like to consider here...

What is actually going to happen to these walls? If they are going to be manipulated in any way, for instance broken down by attacking units or blown up or whatever, then splitting them into pieces would allow some dynamicism.
(is that a word??)

If on the other hand your walls are indestructible (kinda like Syndicate or Doom) then by all means treat them as whatever you like...

I'd appreciate any thoughts on this, and apologize is this is an idiotic question. In building an iso engine I've come to a perplexing problem in dealing with what I call "higher level" objects or things that lie on the ground layer. For instance, walls, trees, etc. The problem is 1)how to make sure they are drawn in the correctly in respect to location and 2)dermining their effects on the walk map. The two approaches I've seen are static walkmaps where the walkmap is hand designed for each map and or image or 2)making sure that objects, like a long wall, are "cut" into pieces that would fit in the location tile. Whew, hope this makes sense!
For example, take a wall that's 3 location units long. I don't want to have to splice it into three pieces but then if I don't the task of determining where in the bmp the start of the wall really is and exactly what the walk locations in the walk map become (as a result of using this large wall bmp), an interesting problem to say to the least. Another example is a tree. Something like an oak may span a large portion of the screen with its broad branches but perhaps only is "anchored" at one location tile. I'm like to hear what others have done to handle this. Thx :-)

Thanks for input all :-).

Andy: that's in interesting idea. I thought about doing something similiar with my terrian by manipulating the vertices (my ground is true 3d). Right now though, I'm just to get the engine up and running I was going to make them static.

While rather archiac I think what I will end doing, to allow me more freedom in image size and definition, is wrapping the image in a wrapper that includes things like the anchor point and walkability. I will have to define those values for each image but it will allow for a lot a art flexibility I think.