Kinesis isn't a good idea for low-latency queueing. It can handle high throughput, but it can often take anywhere from one to ten seconds for a message to make it through the queue.
Given that DynamoDB can reliably write in the 4-5ms range, a kinesis queue may not be necessary. Unless the point of the Kinesis layer is to keep the cost of DynamoDB provisioning low?
It was awful, actually several "wings" in one building. It was possible to get locked out of the building on a walkway between two wings, with no other exits. You would sometimes have to go through one person's office to get to another office.
The theory when I was a student there was that university construction in the late 60s was obsessed with dealing with civil unrest - psych buildings with riot-proof windows, admin buildings with confusing layouts, an open square covered in bricks that get slippery when hosed down, etc.
I could see Amazon using this as a NYC media presence, as much as a store. They could do readings by authors, launch events for Amazon original video content, Kindle product launches, and so on.
Also, six foot HDMI cables. Everybody needs those.
They're a bit vague when describing the "I-was-going-to-buy-it-anyway" problem, but it sounds like they're mainly talking about PPC ads on search engines stealing from SEO links on the same pages. (Some of it seems to be talking about stealing sales from the direct channel, which I have no idea how to handle.)
If your website is already tracking ad performance on a per-ad basis (i.e., tracking all the clicks on that ad, and keeping what you pay per click below what you're getting per click), then optimizing for only the paid search traffic that's incremental, and not cannibalistic, is pretty straight-forward, conceptually.
For a given ad's keyword, instead of optimizing only for the results you get from PPC traffic, you treat your PPC bid as an input that affects sales across PPC and SEO traffic for that keyword, and optimize accordingly.
There are a few extra steps and decisions that need to be made before full implementation, but this is the basic idea. It's actually not too bad to account for.
There are other reasons to advertise in search engines on your branded keywords.
For starters, you have account wide and campaign wide quality scores which impact how much you pay per click across your entire account. Having branded keywords which will have high quality scores can help lower all your costs.
Secondly, your competitors can advertise on your brand even if its trademarked. Owning the top position is cheaper for you and a good idea.
Third of all, You would think a brand would want to own as much of the search results page as possible. It will drive up clicks even on your organic listings.
There are also many case studies that show paid search for branded keywords getting more traffic than just organic listings.
I'll search for a product and amazon all the time. I would just easily buy that same product from Walmart if they had it for less.
Branded search does not mean they will buy from you anyway. It could mean that, but it certainly doesn't guarantee that.
I think this study fails to give serious thought to the phonetic differences of the left-side and right-side letters, and the possibilities for co-correlation there.
The right side has 4 out of 6 of the vowels and 3 out of 4 of the other sonorants, which could easily be argued to be the more sonorous or pleasing sounds in a word. 7 out of the 11 right-hand letters fall into one of those categories.
By contrast, the left side is only 3/15 sonorants. Further, the left side has all the sibilants and a strong majority of the stops.
Pointing out that right-side letters correlate with agreeability is an interesting finding. Claiming that which side of the keyboard the letters are on is the _causal_ factor for agreeability strikes me as lazy.
I completely agree, it's interesting how many people love to jump to economic theorizing about how people will act, and how few people look at the actual cost to implement.
I read through a breakdown in one of the Basic Income books, of what it would cost just to get a $10,000/year basic income. Here's just some of the changes that would be needed:
* cut military spending by a third;
* get rid of ethanol subsidies;
* get rid of mortgage interest tax breaks;
* get rid of "married filing jointly/separately" filing status, make everyone file just for themselves
And that's just the ones that jumped out at me. Virtually all tax loopholes would have to be gotten rid of, and more. All for just the $10K/year basic income level.
So, just as a question of the political cost to implement, Basic Income seems like a laughable non-starter.
There's really nothing new about hostility to immigrants in the United States. The last generation's wave of newcomers is the next generation's "old guard", saying that the new wave is "lazy", "unskilled", and a leech on society. What's going on today seems like a fairly straightforward continuation of how we've always been.
There've always been lots of people who would say that the next wave of immigrants is unneeded, and unwanted. No matter how successful they turned out to be.
This is becoming my biggest concern about these Basic Income plans, too.
Giving this amount of money to every person in Switzerland would total almost half of the country's GDP.
Even if you exclude children and get the total amount down around 30% of GDP, you're still right around the total amount of money that the Swiss government takes in in a year.
So to get this basic income with no new taxes, the Swiss government would have to abandon all its other obligations. Or, put another way, to fund this and still function at the same level, the Swiss government would have to double its tax income.
Remember that everyone above median income is going to end up paying what they receive right back to the government in taxes. Only half the adult population will actually see this money as a check and not just a tax break.
Given that DynamoDB can reliably write in the 4-5ms range, a kinesis queue may not be necessary. Unless the point of the Kinesis layer is to keep the cost of DynamoDB provisioning low?