[mapserver-users] querying WMS layer / WFS reprise

bartvde at xs4all.nl bartvde at xs4all.nl
Thu Jun 3 07:34:03 EDT 2004


Hi Gregor,

perhaps others can comment on this more thoroughly.

I don't think this is a problem of Mapserver's WFS implementation.

I think the performance bottleneck is in the GML parsing part (WFS
Client), and not so much in the GML generation part (WFS Server). Also
ofcourse the transport / download of GML can be time-consuming. Gzipping
this data will save a lot of time on the download (and you can do this
with mod_deflate in Apache e.g.), but I don't know if Mapserver WFS client
can decompress this. The slowness of parsing GML has resulted in looking
at faster alternatives in the OGC world like e.g. binary GML or feature
streaming.

There is not much point in serving your data through WFS, if you are gonna
end up just displaying it through WMS. The only reason for this should be
a distributed system in which you cannot access the data in any other way.

Best regards,
Bart

>
> On Thu, 3 Jun 2004 bartvde at xs4all.nl wrote:
>
>> > Is this cascading thing an issue for MapServer's WFS implementations
>> > as well? Could a MapServer WFS-client layer be queried?
>
>> I just tried this and you can query a LAYER of CONNECTIONTYPE WFS. So no
>> problems there, but that's because this is a GML "datastore" just as
>> shapefile is a "datastore". With CONNECTIONTYPE WMS you cannot access
>> the
>> data itself as a (WMS) client, that's the big difference.
>
> Excellent, that's what I would have expected, since it's just vector
> data by the time the client/displaying MapServer sees it. Thanks for
> confirming that for me - you saved me an hour or two.
>
> I spent the last day experimenting with WFS, since that's what we *really*
> want to be sending instead of images, and came to the conclusion (that
> seems to have been confirmed by some others on the list) that WFS is
> incredibly slow (as in, several minutes) because of the GML encoding,
> which made it unsuitable for use in dynamic map layers.
>
> WFS' slowness surprised me: should a feature list being sent to the client
> never contain more than several items? Or is this a "problem" with
> MapServer's implementation of a WFS server? Would another WFS server
> (GeoServer?) do it faster and be able to send more items in a reasonable
> period?
>
> _______________________________________________
> Mapserver-users mailing list
> Mapserver-users at lists.gis.umn.edu
> http://lists.gis.umn.edu/mailman/listinfo/mapserver-users
>




More information about the mapserver-users mailing list