From having spoken to people who worked on object clustering for Froogle in the past, I think the base issue is that it's surprisingly hard to extract all the metadata accurately from descriptions. Even if you provide an file format and allow merchants to upload all their item descriptions with structured data, you'll still just end up getting bad data from a lot of merchants, and then you have the problem of whether you want to accept the inaccurate data provided by the merchant who has an incentive to misstate things, or the inaccurate data provided by your flawed automatic text extraction systems.

I'm not sure if there's a good workaround for this besides buying from places specializing in the specific types of items you're looking for, which comes with its own set of tradeoffs in price, convenience, etc.