[OpenLayers-Dev] Reasoning behind Layer.HTTPRequest selectUrl() logic?

[Paul Spencer]

On 12-Nov-08, at 2:39 PM, Christopher Schmidt wrote:

> On Wed, Nov 12, 2008 at 01:59:59PM -0500, Nathan Gerber wrote:
>> A member of my team and I are curious about why the selectUrl()  
>> function
>> within Layer.HTTPRequest is as it is? It seems like alot of  
>> unnecessary
>> overhead to pick a server to submit your requests to. It looks as  
>> though
>> alot of thought went into this but it just doesn't seem to make  
>> alot of
>> sense to either of us. It seems to parse through an entire  
>> parameter string
>> which should be identical for each tile aside from the bounding  
>> box. It also
>> seems that if I have an array of 3 servers and one of the servers  
>> fails when
>> I do the retry attempt it will go right back to the server that  
>> failed
>> instead of attempting to make use of another server within the array.
>
> I think the answer, in a single wrd, is: "Caching."

also (although of less practical value to most people), if there are  
multiple instances of tilecache (or presumably some other piece of  
software) that manage the actual tiles on disk, the tiles associated  
with each instance are only on that instance.  This was an issue for  
me using tilecache on amazon's ec2 because of limited disk space per  
server, so I set up multiple servers that each handled part of the  
back end tile cache.   If OpenLayers could send the same bbox to other  
urls, I would have had a problem because the tiles would have been  
duplicated on each machine (and rendered by each machine too, more  
overhead).

Cheers

Paul

__________________________________________

    Paul Spencer
    Chief Technology Officer
    DM Solutions Group Inc
    http://research.dmsolutions.ca/