feedback on possible mapserver enhancements

[Howard Butler]

On Feb 5, 2008, at 1:24 PM, Stephen Woodbridge wrote:

> Daniel Morissette wrote:
>> Steve Lime wrote:
>>> Good suggestions Arnulf. I too think a WFS-T-lite makes good sense.
>> Without having done much research, I like that idea as well.
>> Daniel
>
> I would like to add my +1 to having some type of WFS-T support  
> within mapserver.
>
> My basic use-case would be something like using OpenLayers with  
> mapserver serving images, and want to be able to support simple  
> point, line, polygon, box, circle like objects that can be  
> displayed, created, edited using OpenLayers vector support, and  
> saved these back to a postgis database?


> without the need to install and manage any additional major  
> applications, like geoserver, featureserver, etc.


This seems totally silly to me.  FeatureServer is a simple cgi  
application just like MapServer.  Drop it in your cgi directory, tweak  
a configuration (mapfile), and be on your way.  No code to write and  
no datasource limitations (MapServer's wfs-t likely will only support  
one or two drivers, not all).   In fact, you can do this right now  
without compiling *a thing*.

What is the driving factor for having WFS-T in MapServer?  Is there a  
funder with some really, really deep pockets and strong commitment to  
make it happen?  Is it a "gee, that would be nice" sort of feature  
with a couple of folks musing and a tiny bit of funding to take a  
crack at something?  Is it really related to the fact that you can't  
share configuration between MapServer and any other application server  
in the GFOSS domain and people want simple/single configuration ability?

MapServer has no history of doing anything like WFS-T AFAIK.  We don't  
do schema validation, we serialize GML with printf for chrissakes, and  
we don't have consistent behavior of our data drivers.  I'm not at all  
convinced that we can do right with WFS-T and end up with something  
that's less than a PITA of installing another software that is  
designed from the ground up to be doing WFS-T.  Use the right tool for  
the job is many folks' mantra, and IMO, MapServer isn't going to be  
the right tool for this job even with WFS-T-lite.

I'll gladly eat crow and wait to be proven wrong when MapServer comes  
out with a bulletproof and complete WFS-T implementation that  
interoperates with everybody and shines as *the* reference  
implementation of such a thing.  GeoServer has *years* of development  
invested in WFS-T.  Are we prepared to do the same?

Howard