[pycsw-devel] CSW Distributed search question

[Tom Kralidis]

Hi Jose: comments interleaved:

________________________________
> From: jose.garcia at geocat.net 
> Date: Thu, 2 May 2013 16:16:36 +0200 
> To: pycsw-devel at lists.osgeo.org 
> Subject: [pycsw-devel] CSW Distributed search question 
>  
> Hi all 
>  
> I'm checking to implement DistributedSearch for GeoNetwork and the OGC  
> specification seem not very complete related to how to return results.  
> I have check pyCSW as it's implementing it to see how has been managed  
> and seem like this: 
>  
> - Supposing local catalog and a federated one. 
>  
> - A GetRecords query is send with startRecord=0 and maxRecords=10: 
>  
>     - Local node matches 12 results, so should return the first 10. 
>     - The same request is send to the remote catalog, that matches 24  
> results, so returns recordsMatched=24, recordsReturned=10 
>  
>    - Local node aggregates results and returns: recordsMatched =36,  
> recordsReturned=10, nextRecord=11 and returns local and remote records  
> (20 in total) 
>  
> I understand why returned the total records (20), otherwise when  
> requested startRecord=11 and maxRecords=10, if only returned 10 in  
> first request, the other 10 will be not returned in any way. 
>  
> Did you find about this behaviour in some other spec document or you  
> just decided it was the way to go? 
>  

We just decided it was the way to go in the absence of clarity from the standard.  As well, our initial implementation of this did/does not want to go down the route of resolving relevance of records across nodes before returning to the client.

Having said this, CSW 3.0 addresses this issue with returning GetRecords w/ DistributedSearch results grouped by federated catalogue so it will be easier to represent this when CSW 3.0 comes out.

Let us know how GN proceeds with this, as suggestions would be valued.  It would be great to have this functionality consistent across CSW 2.0.2 implementations.



> Also not a big expert in python, so hope no misunderstood the the code  
> and also the problem described is clear. 
>  
> Thanks in advance and regards, 
> Jose García