[postgis-devel] [raster] Paths local to server

[Mateusz Loskot]

On 28 February 2013 21:01, Bborie Park <dustymugs at gmail.com> wrote:
> On Thu, Feb 28, 2013 at 12:58 PM, Mateusz Loskot <mateusz at loskot.net> wrote:
>>
>> On the course of discussion about ST_AsBinary binary [1],
>> it's been revealed that local paths play important role as
>> part of public protocol and API in PostGIS Raster.
>>
>> If that is true, then this is a mistake and design flaw.
>> The WKB format in RFC2 does not consider public API
>> and Sandro explained to me it does not thorough consider
>> out-db storage mode.
>> It was clear to me from the very beginning, that the isOffline flag
>> is for internal use only, so we can optimise parts of internal
>> implementation.
>>
>> Existence of utility functions like ST_BandPath is not enough to
>> support further spread this mistake onto other, more crucial,
>> parts of the SQL API.
>>
>> I'd just like to confirm if this 'local path' madness is going to settle
>> for good in PostGIS Raster API.
>>
>> [1] trac.osgeo.org/postgis/ticket/2217
>
>
> I think the problem really is that the WKB format is flawed.

Flawed may be unfair, as the first part of it is fine and well-defined,
I'd say it this way (as discussed on #postgis):

<mloskot> There apparently is empty page left in RFC2 and some things
are not complete there. Later, we took this as completed work, without
reflection
<dustymugs> yep

And that is exactly what Sandro seems to confirm as well.

> That format (as adopted) doesn't work for in-db versus out-db.  WKB
> shouldn't care if it is in-db vs out-db. Just that it provides a
> consistent output where band data is provided.


Best regards,
--
Mateusz Loskot, http://mateusz.loskot.net