Can ArcGIS Server Ever Really Replace ArcXML?
Those of you reading this that know me, know that I haven't been a big advocate of ArcGIS Server. I dream in ArcXML. Really. It will be a sad day when ESRI retires ArcIMS. That is, if they really do retire it.
That being said, I have seen ArcGis Server come a long way since that initial 9.0 release. 9.3 shows lots of promise, and I am especially impressed with the new javascript and REST api's. For most folks out there that have to produce apps on a regular basis, the web adf has been a disaster. Its simply not easy (compared to ArcIMS/ArcXML) to create complex, on the fly maps. Period. I dont think there is much argument to that point.
My coworker came in my office this morning and said "What do you think about migrating the xxx app over to ArcGIS server 9.3 stuff you have been crowing about?" My first thought, was yeck yeah, that would be cool. Its just some simple maps displaying some simple geography with nothing fancy going on. Perfect fit for the new 9.3 Javascript/REST API. But, then it occurred to me...to make those 'simple' maps, we are doing some very tricky ArcXML to get the right data displayed on the fly. You know they type - we have all done it - something that really pushed ArcXML to do something very cool. Its probably a lot more than the creators of ArcXML ever intended.
Then I thought about it more...how in the world would I do this with ArcGIS 9.3? Use the web adf to display a simple map? Nope. Not a chance. Use REST? Well, its nowhere near ready to take on these types of crazy requests that we can do with ArcXML (like on the fly rendering, etc). In the past, we could massage that ArcXML to do just about anything we wanted. The downside was we didnt get that instant gratification of a ajaxy map that AGS 9.3 gives. The general public now demands the ease of use and speed that The Google gives us.
So, where does this leave us? In a tough spot. AGS still isnt ready to create those cool on the fly maps (not everything can be cached you know!) that we could do with ArcXML. My hope is that the ESRI REST implementation will continue to grow, as I'm sure it will. But, will it ever be able to achieve that ArcXML status? My thinking is that ESRI is going to keep ArcXML around for a long time to come and it will be integrated into future releases of ArcGIS Server. There must be some middle ground here. I'm sure they have already thought about it plenty, but now is the time to act.
But, I really should have stressed just how happy I have been with 9.3 and the js/rest api. I think ESRI understands how important it is to get this thing right and continue to extend its capabilities. If I were you, I'd hold tight and wait for 9.3, its pretty awesome.
I've had some very good results with using ArcIMS to render maps with data computed on the fly using a variety of front-ends to let the users interface with the maps. My favourite is, of course, an in-house Coldfusion and Javascript AJAXy interface that's trivial to customise for each new application, but we've occasionally used it against other HTML interfaces such as the Google Maps API as well as the usual Standard machine readable interfaces - again with a CF wrapper to apply whatever logic is needed.
I think ESRI would be onto a winner if they kept ArcXML and IMS vaguely up to date with some of the more recent rendering features. Supporting more than one fundamental approach to automated map generation will clearly satisfy more of their paying customer base than, say, dropping IMS and just pushing Server. I know if that were to happen, we'd just switch to something like Mapserver, there are many people in the world who enthusiastically dislike the 'visual studio' style approach to writing web systems.
Now, where's the official ESRI CFC version of the CF Connector :)