[Tiling] compact tile format anyone

[thomas bonfort]

Mateusz,
I've rapidly read the wktraster spec and there seem to be a few
shortcomings that make it seem incompatible with the requirements for
a tiling storage. imho the spec would have to evolve on these
following points to make it a usable implementation for tiling
requirements:

* raster data is either stored uncompressed in the database itself, or
as references to external files: with the actual spec we are either
left with having to recompress data to png/jpeg on the fly which is
incompatible with the high performance needs of a tiling server, or
having to reference external files which does not solve the problem of
a self contained compact tile format. could the spec be enhanced so
that a wktraster row can contain a "blob" that's the actual compressed
image?

* no pyramids (reduced resolution coverages can be stored as a
separate layer): as a tileset is an actual pyramid, storing each tile
resolution as a separate layer doesn't seem like a very elegant
solution.

* no multiple dimensions (eg. wms time): probably less important, but
it would be nice if the format could take into account multiple
dimensions natively, as it is not an uncommon tiling requirement.

best regards,
thomas


On Fri, Dec 24, 2010 at 13:43, Mateusz Loskot <mateusz at loskot.net> wrote:
> On 24/12/10 12:29, Oliver Tonnhofer wrote:
>>
>> On 23.12.2010, at 12:19, Mateusz Loskot wrote:
>>>
>>> I'd suggest to take a look at specification of the PostGIS RASTER
>>> (formerly known as WKTRaster) and see if it would fit the ideas
>>> and needs discussed here. If something is missing, I believe the
>>> spec can be improved and PostGIS will benefit as well. Once the
>>> spec is made "portable", it's a matter of implementing it for
>>> SQLite (perhaps with collaboration of SpatiaLite).
>>
>> I haven't had the opportunity to play/work with it, but WKTRaster
>> sounds really interesting. It does much more than we need for a tile
>>  storage and so I think it would be too large/complex dependency.
>
> Oliver,
>
> I am not suggesting to use WKTRaster implementation, so no actual
> dependency.
>
>> I'll have a look at the spec and see if we can share at least some
>> parts so that it would be easy to use PostGIS Raster as a tile
>> backend.
>
>
> Yes, I am suggesting to look at the WKTRaster specification
> which is, in my opinion, well crafted to handle variety of tile
> configurations and schemes.
> Such collaboration on the spec level potentially could be beneficial for
> both sides and would make WKTRaster a good test bed and reference
> implementation.
> Not to mention that if same schema for metadata is used by various
> implementations for server dbms and file-based databases, then
> interoperability and portability is increased. Think of OGC standard.
>
> Best regards.
> --
> Mateusz Loskot, http://mateusz.loskot.net
> Charter Member of OSGeo, http://osgeo.org
> Member of ACCU, http://accu.org
> _______________________________________________
> Tiling mailing list
> Tiling at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/tiling
>