[postgis-devel] Motion: vote for considering SRID <= 0 as "unknown"

[Chris Hodgson]

On Tue, Oct 04, 2011 at 10:20:51AM -0700, Paul Ramsey wrote:
>> For the difference between "implementation-defined" and
>> "implementation-dependent"
>>
>> http://www.lispworks.com/documentation/HyperSpec/Body/26_glo_i.htm
>>
>> Basically the spec is saying that 0 is the only "special value"
>> allowed, and we can say what its specialness is.
>>     
I actually interpret it as saying that zero MUST have a special meaning, 
which is defined by the implementation. Other values MAY have special 
meanings. So that would allow us to get away with <= 0 as "special" == 
unknown.

Strk I don't think you can separate these issues, they are 
interdependent. The issue is standards compliance vs. legacy 
compatibility, and we can satisfy both to varying degrees if we consider 
all of the alternatives in concert. That said..

It seems to me that if we want to switch to zero, allowing <= 0 as 
unknown will allow for -1 inputs. I'm not sure that anyone really checks 
for -1 outputs? Really, this gives us the best backward compatibility, 
and provides a simple change for external applications that will also 
support backward compatibility (for users who didn't make custom SRIDs 
with negative values, and I'm ok with that). We could also use this to 
provide a migration period, whereby we could actually switch to 
returning 0 instead of -1 at some point in the future, after we've given 
client apps time to make the change to supporting <=0.

So I'll +1 this.

Chris