Hi Everyone,
I'm working on my first IDX / MLS project and figured I'd post here before I got too deep into it.
The initial project is to provide live enriched listings for a couple of real estate firm clients. It's going to start off small with just live for a few realty sites listings and eventually grow to encompass a lot more.
I'm deploying on one of my dedicated Windows 2003 servers. Probably will use Microsoft Sql Server 2005 for the database and the application will be written in .Net/ASPX/C#
I'm receiving FTP feeds from North Florida MLS and I wrote a job to uncompress them into data CSV files.
The first issue is the data structure itself. The schema supplied by NFMLS has a huge number of columns in each of the data structures. 300+ in the smallest one. I'm used to working with highly normalized schemas with a lot of referential integrity. The data feeds I'm getting from NFMLS are giant denormalized basically flat files. Don't know what solutions you've come up with but tables with hundreds of columns don't seem like a good idea. And in the case of Sql Server they exceed the max row size by a fair bit. I'm not opposed to moving to MySql if that database would handle these huge row widths better.
The client isn't going to want to spend the money required for me to normalize this whole thing so I was looking for a simple fairly quick solution.
My one idea was to pre-parse the MLS data and cherry pick the 50 columns of interest into a tighter table containing the information that you'd need 90% of the time and then maybe stream the other stuff into XML and stuff the XML doc into a blob and then run XPATH queries on the XML DOM to present the more obscure stuff (like if you can waterski behind the place or other nonsense like that)
The only problem is that this solution boxes me into the handful of columns that I pick as important right off the bat.
If anyone out there has any ideas I'd love to hear them.
Thanks for your time.

Reply With Quote

