Glad you think so highly of the presentation, but I don’t have the raw files so can’t put the video up anywhere else.

Thanks.

Any place we can get a better viewing experience for this most excellent of presentations? On every system we’ve viewed this on the Vimeo player appears to lack (or restrict) full screen playback, fwd/rwd/skip features, and audio control (though of course the big volume dial on my speakers does mitigate that last one +D).
So in recap,

Presentation: Awesome
Vimeo: For Suck (seriously it’s complete schlock…)

It appears to go in and out – not sure why.

Real-time data just has tighter staleness requirements – under 5 seconds, under 1 second, under 50 microseconds for high-frequency trading.

Reading through your blog post, I didn’t see you addressing the trade-offs between how often some thing happens versus the cost of preventing/ensuring it happens 100% of the time.

The question/thinking processes that I describe are the important parts of the process, not the answers that I use to demonstrate them.

Thanks for your comments.

I can agree that the name CustomerMoved sounds more like an event than a command, and does not capture the context of *who* is doing it – which matters to whether the command is accepted or not.

Naming is hard, as it drives to the heart of business modeling.

Thanks for your comments.

http://www.softec.org/blogs/central_coast_code/archive/2010/03/09/beware-stale-data.aspx

Thanks again for your time in putting together presentations like this for us to watch.

I think I phrased my question wrong, I knew what has to happen on the technical side, i. e. send something that contains the userid and the product id.

I think I just don’t like the phrase command in this case as it is not really a command, all other command are more like MoveCustomerCommand and not CustomerMovedCommand.

Anyways I think I’ll just accept some “badly named” commands to keep things simple.

Thanks again.

I hear you. I ran into that with Ayende’s presentation as well. Will check with Skills Matter.

In the Amazon example, on top of the query, it sends a command – CustomerViewedProductPage to the back-end with a session ID. When someone actually buys a book, the purchase contains that same session ID allowing for correlation.

In short, since a command is a fire-and-forget affair, you can easily add it to those scenarios and handle it using standard CQRS.

Does that answer your question?

I did take a cursory look at it. I think that it shows the use of some technologies that may be involved in a CQRS approach, actually focusing more on Event Sourcing.

One of the important elements of CQRS as presented in the video is the use of the Query model for preliminary validation before commands are sent – that is not demonstrated in the sample.

Hope that answers your question.

Thanks.

There is one thing about CRS that I don’t get, how do you handle events that are the outcome of a query?

If you’d do something like Amazon’s xy % bough the product shown on this page, zx % bought product yy…. etc, how would you do that with CQRS?

That would mean, each time a customer views a product (thereby issues a query) an CustomerViewedProductEvent has to be fired. This is not a command as it is a fact, it already happened.

How would you get that into your domain / event store?