Edit: Once while sitting on the private sector side of the table (yes RFPs were often drafted in consultation with the company that would eventually win them, if just informally), a [state] government employee expressed concern over the wording of a particular requirement because he was worried that it would open the door to our primary competitor being qualified to fulfill the bid.
RFI process is usually far more open.
Having said that, just as often, it's due to cronyism.
The line between cronymism and efficiency is in the eye of the beholder.
It's pretty clearly an example of corruption; the entire purpose of single-sourcing a contract is to defeat the RFP, or bidding-out, process. A decision is being made by someone who isn't supposed to have the authority to make it.
"Defeating an RFP", then, is frequently (behind closed doors) agreed to be in the best interest of good government. A chosen expert—despite the bad optics—can do better (usually much better) than the RFP process.
I'm not sure "corruption" is the right word for this, since the goal here isn't to line anyone's pockets, but actually to choose the consequentially optimal, highest-ROI option. It's not quite "two wrongs making a right"... maybe "three lefts making a right?"
Did you type this on the cheapest thing capable of making HTTP requests?
Do you live in the cheapest structure capable of maintaining its interior at a non-life-threatening temperature?
I submit that lowest-bidder rules are patently absurd. People who have to deal with something day in and day out ought to have the authority to select a reasonable one.
Rigging a bid in a way that compromises the real task at hand in order to personally enrich yourself is clearly an example of corruption. Rigging a bid in (what you think is) the interest of the task at hand, but in violation of contracting rules, is rule-breaking and insubordination but not corruption.
It's also, more importantly, opening the door to new markets like banking, insurance and state/local government. Many compute intensive workloads in state/local government remain on-prem beacuse of the lack of Federal guidance, which means higher compliance risk.
Silicon Valley drives innovation, but the federal government drives implementation of technology.
https://cloud.google.com/compute/docs/instances/creating-ins...
a) Have you ever known an IT project, let alone a government IT project, to end up smaller than or equal to the projected size/cost? Or anything near it?
b) If you ever have the “pleasure” of dealing with government IT, you'll see how much easier it is to use a combination of things that has already been “certified” by another part of government. Winning a contract worth 1% of your company's bottom line is huge itself, but as a foot in the door, it's massive.
Your points are all the ones raised in that... wapo? nyt? article that you're alluding to, which was discussed here and found to be largely nonsense, IMO. For example, that 28/30/32GB thing is wrong. Are you really saying GCP and Azure don't offer 32GB options? That's a totally standard and common configuration these days.
The idea for example that MS are effectively excluded from bidding because they can't achieve the feat of putting 32GB of memory in a server is a bit absurd when you think about it. For a big contract like this I reckon MS techs could probably just about manage it.
Have the people drafting the requirements done a bit of copy-n-pasting? Definitely.
I'm surprised this isn't being discussed more. With a huge $10bil cloud I would think you would want to select a secondary vendor to distribute redundant workloads and data storage. Even if Azure or GCP couldn't meet some requirements today it's a ten-year project and both are expanding quickly. Azure has recently brought online government-specific locations and Google has already has geographically dispersed zones within regions that meet these requirements.
Btw, Google is years away from getting the kind of security certifications you'd need to run a cloud from the Pentagon. Oh, and the Project Maven thing.
https://en.wikipedia.org/wiki/Rugged_computer
This is the kind of astroturfing that should get your account flagged.
I'm pretty sure "ruggedized" is a common term for pseudo-milspec computer hardware. You see it on the marketing of regular ol' external USB hard disks quite a lot. If Microsoft and/or Google had a product of the same type, they'd very likely use the same adjective to describe it.
It was discussed recently on HN and, IMHO, mostly discredited. For example, the RAM amounts are wrong for Google and Azure, and certainly they have 150 miles between data centers.
Having said that, AWS may realistically be the only provider that can meet their needs.
But I suspect the 150 mile marker has some other objective driving factor behind the reasoning for that number, which Amazon incidentally has some advanced insight on, due to other commitments.
It probably draws directly from line-of-sight targeting at visible horizons for some sort of specific altitude. Zero altitude, mean sea level gets you maybe a thirty mile shot with conventional artillery, targeting a multi-story building. But then again, rail guns seem to be capable of sending shells or slugs over 200 miles [0], so who knows...
[0] https://www.cnet.com/news/futuristic-navy-railgun-with-220-m...
Some other AWS-only requirements:
* exactly 32gb of RAM. Google promises 30, Azure promises 28. We all know how realistic this isn't. * 150 miles between data centers. This arbitrary number is exactly what Amazon advertises. Others can manage 100 miles, or 130, but this number comes right out of Amazon's advertising package.
Not to mention the idiotic idea of a single provider solution.
On the other hand, this is a 10 year contract worth about 1% of Amazon, Google, or Microsoft's bottom line in one year. So there's a bit of mountain/molehill going on here