MS RFC 22a: Feature cache for long running processes and query processing (update)

[Daniel Morissette]

Umberto Nicoletti wrote:
> Warning: rant ahead
> 

Warning: rant continues... but hopefully that's my last reply in this 
thread which is turning into a debate between those who want a Web GIS 
and those who want to stick to MapServer's original goals of being the 
best and fastest engine for rendering maps.

I think we need to have this debate since it's important in order to 
define MapServer's future direction but perhaps this thread is not the 
right place.

>> > I think the users who need a Web GIS should be looking more at MapGuide
>> > than MapServer.
> 
> I share Tamas' concern about this statement. If MapGuide does it all
> then why am I spending so much time on developing and supporting
> mapserver? Because MapServer is open source, is small, fast and
> efficient 

You answered it yourself: "because mapserver is small, fast and 
efficient"... at what it does best: pushing maps to the Web.

My opinion is that if we keep adding GIS processing type of features 
we're adding complexity and getting away from the initial goal of 
optimizing every line of code to render maps which was MapServer's 
initial goal and the reason why *I* spent so much time on it since 2000. 
More complexity means bloat, more chances of bugs and slowness to me.

That's why I tell people who want advanced GIS features on the Web to 
use MapGuide and I tell people who want a transactional WFS to use 
GeoServer: because I want to keep MapServer small, fast, efficient, 
robust and as bug-free a possible and I don't think those features are 
compatible with my goals.

Am I lazy and trying to avoid the problem of implementing those features 
efficiently? Maybe... but hey, the approach has worked fairly well so far.

I'm not worried at all about the future of MapServer as long as we keep 
it small, fast, efficient and robust. I see MapServer, MapGuide and 
GeoServer as complementary tools and not as competitors. We have told 
people to use GeoServer for transactional services for years and 
MapServer has only benefited from that since we have not had to change 
its internal data structures to deal with that problem... and as a 
result today MapServer is still the fastest at rendering maps. This was 
acknowledged again by the developers of both GeoServer and MapGuide in 
the OSGeo BOF at Where 2.0 a few weeks ago.

If you think I'm making a mountain with nothing, here is a good example 
of an apparently harmless feature that turned into a performance hit:
http://trac.osgeo.org/mapserver/ticket/162

> 
> Maybe all this 'machinery' for caching could be moved out of the core
> and into mapscript so that it can be easily programmed in a higher
> level language?
> 

I think at a minimum the problem of double-pass queries needs to be 
solved for the mapserv CGI as well, so a mapscript-only solution would 
not work for that.

Daniel
-- 
Daniel Morissette
http://www.mapgears.com/