[Rets-dev] 2.0 overhead redux (long)

Matt Lavallee matt at pmptechnology.com
Wed Apr 25 20:20:39 CDT 2007


Alright, without arguing point by point, here are my issues:

1. Bandwidth.  I acknowledge that I am in a very large MLS by comparison, so my
payloads are equally large compared to the average MLS/client transaction.
However, bandwidth is by no means free to anyone, and I doubt very many MLSes
are equipped (or eager to hear about) a potential 30-50% increase in their RETS
utilization.  My MLS (again, a high-water case) has two bonded T3s with very
high utilization, and it is *not* a trivial cost to add another T3.  We (one of
many client services in the area) burst against their pipes at 30Mb+/s.  Also, I
heard talks of systems handling _millions_ of RETS transactions a day... even
with a 30% (being generous) increase in overhead, that is not a marginal
expense.  And "several-fold" means "several times as many" according to
Webster's Dictionary.

I didn't even mention the metadata sections because the metadata dictionary
structure is similarly much, much larger than that of RETS 1.7.  While trivial
in large payloads (100+ listings), it will be significant in small payloads
where the metadata is the bulk of the transaction.

2. Data Model.  This is not a topic I can delve into in any pleasant way, since
I tried to raise the point more than once on-list and got no response...  When I
see BathsTotal, BathsThreeQuarter, BathsHalf, and BathsFull, I say that it's not
very modeled.  Ideally, the presentation would be:
<baths>
	<bath type="threequarter" level="0" />
	<bath type="half" level="1" />
	<bath type="full" level="2" />
	<bath type="full" level="2" MasterBath="true" WainscotMaterial="ceramic"
/>
</bath>

... and "BathsTotal" would be a derivative formulation of the above.  Moreover,
the semantics would be consistent and TotalBedrooms would appear with the same
<object><function> notation.

I could go on and on about the model that includes ListingProperty, Listing,
Property, PropertySummary, ListingHistory, and ListingRecord being disparate
schema describing a single entity, but I _really_ don't want to talk about the
model.

-Matt





More information about the Rets-dev mailing list