[postgis-users] MapServer with PostGIS: motivations

Attila Csipa plists at prometheus.org.yu
Mon Aug 28 15:26:06 PDT 2006


On Monday 28 August 2006 22:33, Brent Wood wrote:
> Apologies for this second post. A couple of aspects I forgot to mention :-)

Just a few observations:

At the point where a mapserver/postgis gets to be a distributed system, there 
will be many factors in play which influence both performance and 
manageability. You might run into network bottlenecks, and also the type of 
data will influence how much you actually benefit performance-wise (your 
bottleneck might be Mapserver itself). I lost count how many times I heard 
something needs to be done with/through a DBMS instead of simple binary files 
since 'it will surely be faster'. That is plain wrong, and definitely case 
specific, but let's not get into the discussion what's the original purpose 
of databases :). What is REALLY good on a database is generally not 
performance but ease of editing and operations on both geometries and 
attributes. Imagine what happens if you have shapefiles on a file sharing 
scheme and you want to edit a record, insert or remove one, etc. All the 
transaction mechanisms, locking, etc, you would need to implement on file 
level is already done, and far better for this purpose, on the DB level. 
Dynamic layer content is death to shapefiles. Shapefiles are not that new, 
and the reason they are still around (well, apart from inertia and 
non-technical factors) is that they are simple and they do what they do 
fairly well, apart from some historical shortcomings. They're not complex, 
they're not smart, they were not intended to be that. You can always think up 
a task where a particular spatial data storage will come up first. Also, many 
things you mentioned CAN actually be done with shapefiles with regard to 
mapserver. You mention simplify, and class separation - if you have a high 
volume of requests it pays to have specific shapefiles pregenerated if you 
know your specific dataset will benefit from it. All in all, among the many 
valid and many YMMV reasons to use postgis over shapefiles, raw performance 
would be pretty low on my list, especially if I had little to no experience 
with postgis. Postgis is not a magic bullet, but a very nice tool - use it as 
such :)





More information about the postgis-users mailing list