[Tiling] Fwd: [OSGeo-Standards] Irregular tile coverage in TMS?

[Matt Priour]

In most instances, if you were having your tile creator cut tiles for only 
where data was present, then you would end up with the same type of pattern 
as the red example. The actual definition of the tile set would be the same 
as the regular tiles covering the entire bounding box of that tile set. It 
is then up to your tile server as to what happens when it receives a request 
for a non-existent or empty tile. I think requiring servers to actually 
cache individual empty images for each empty tile is not useful.  The server 
response, whether it is an no data image for each request, a redirect to a 
common no data image, or a simple 404; is not really all that important 
either. Clients should be able to handle any of those type responses and 
behave appropriately.

Matt Priour
Kestrel Computer Consulting


--------------------------------------------------
From: "Mateusz Loskot" <mateusz at loskot.net>
Sent: Sunday, September 12, 2010 3:13 PM
To: "Oliver Tonnhofer" <olt at bogosoft.com>
Cc: <tiling at lists.osgeo.org>
Subject: Re: [Tiling] Fwd: [OSGeo-Standards] Irregular tile coverage in TMS?

> On 12/09/10 19:54, Oliver Tonnhofer wrote:
>> On 12.09.2010, at 13:33, Mateusz Loskot wrote:
>>> I understand the size and spacing constraints. What I'm missing,
>>> however, is what TMS says about "topological" regularity of tiles
>>> within a coverage.
>>>
>>> I attached basic drawing with two example tile coverages:
>>>
>>> a) blue - regular coverage, rectangular
>>>
>>> b) red - irregular coverage
>>>
>>> Is only blue version kosher for TMS or both versions are valid?
>>
>> It should be valid if your grid size at this level is 4x4 and if you
>> are returning empty/transparent images for tiles outside of your
>> coverage.
>
> Yes, transparent, empty or tiles of some NODATA rasters...
>
>> An optimization would be to return 301/302 and point to an
>> empty image with long expire time, but I don't know how redirects are
>> supported in non-browser clients.
>
> This is a Web client aspect which is beyond my question.
> I'm trying to discuss TMS without considering
> on client side.
>
>> You've got the same problem with WMS/WCS where you can only define a
>> bounding box in the capabilities. Requests outside of the coverage
>> are transparent, or only contain the background color.
>
> This one gives important input to my question, indeed.
>
> Thus, I'd update asking about actual tiles: are those empty/transparent
> tiles actually stored/cached by TMS provider?
> If they are, then physically stored tiles coverage could be assumed
> as always rectangular.
>
> Best regards,
> -- 
> Mateusz Loskot, http://mateusz.loskot.net
> Charter Member of OSGeo, http://osgeo.org
> _______________________________________________
> Tiling mailing list
> Tiling at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/tiling
>