Barrack's page

This is some monday morning quarterbacking here but unless I missed something I think it needs to be said.

I may have missed it somewhere but from what I read it seems as if there is a huge design flaw in the system. If each click of the download spawns a new request despite an existing identical request being in the queue then you have an easily abused setup especially since you have free downloads. You have no way of knowing that a user is who they say they are and they are quite capable of bringing down you servers even after this current issue ends. That has to be fixed in a better way than asking users to read this thread and please stop doing it. The user id and product id seem like a unique enough id for indicating the personalization status (empty, in queue, ready, expired are at least some of the states) Then you can at least report back to the user the state and keep new requests from bogging down the system.

Another possible option is to just turn off on demand personalization for humble bumble products for now. As a batch job personalize all the humble bumble products, store them on an Amazon S3 instance and retrieve them. As it is you're paying for computers to be running full tilt all day and for the most part it seems getting nothing for it. The user disappears before they can download it and the personalized product doesn't exist in the storage server because it's booted out for the next product by the time a user comes back.

It just seems to me you need to stop the whole show until something gets fixed in the architecture because this is just a waste of resources and everyone's time.

I'm intending this as constructive criticism. I don't envy anyone in this position and I've personally been in similar ones. It's no fun and I don't think bad decisions were made. The system worked until now. Now you're getting flaws exposed and some appear to me have impact beyond the immediate load spike from a successful marketing job.

My 2 cents.

Watermarks are a form of copy prevention by shaming and limited tracking. They don't prevent copying they just make it easier to figure out who released it into the wild. That works best when the overlap of people interested in the product, the people interested in stealing it, and the people technically capable of removing the watermarks are small. The larger the overall population getting the product the more likely you're going to get an overlap that results in a cleanly watermark stripped copy out in the wild.

It seems to me it might be worth just seeing if the predicted lost revenue from those books out in the wild is greater than the cost of new hardware. If it's not you might just turn off or severely lighten the watermarking on the humble bumble books. Otherwise you're spending money but getting little for it. The honest folks will still pay. The dishonest folks will not.

The watermarking scheme for your usual sales seems like a perfectly reasonable tradeoff but I suspect at this level it's less successful.