You might want to focus on jet fuel, since we don't know how to make batteries with sufficient energy density for long range flights and don't know if such energy density will ever become possible.
It might also be worth looking at where airplanes tanker fuel, that is to say, carry more fuel than they need for their current leg because refueling at the next stop would be difficult or expensive. Apparently a lot of that currently happens on short flights to small islands; while I hope Wright Electric and/or the EViation Alice will eventually take over that market, in the short term that's a market that might be willing to pay a bit more for liquid fuel made from air plus local solar panels.
Fuel oil that can be burned in combined cycle power plants in the winter may also be valuable for dealing with seasonal imbalances in demand vs renewable generation that lithium ion batteries can't cost effectively balance.
If you can do validation from a perspective where your ISP can send you a full BGP table, you should be able to keep track of how recently the most specific route to that destination has changed; something that has changed recently or that currently has an AS path different than what has historically been typical may be suspect.
AWS does have some network diversity in different regions; they apparently do not operate their own backbone to interconnect the different AWS regions.
(And for all that Hurricane Electric seems to have not been doing a great job of filtering BGP announcements from peers, they do seem to have pretty aggressive pricing on colo space in their Fremont data center where they can provide a full BGP feed.)
FLAVOR: Ubuntu Desktop, Ubuntu Server, and Ubuntu Core
HEADLINE: traceroute
DESCRIPTION: Installing some version of traceroute by default may be desirable, because sometimes when you find yourself wanting traceroute, it's because you want to debug a problem that happens to prevent installing packages over the network.
If I try to run traceroute on a system with no traceroute package installed, I get a message telling me I can either install traceroute or inetutils-traceroute. It doesn't explain what the tradeoffs are. It doesn't explain why Ubuntu can't simply have one good traceroute program that does everything.
mtr can also be good, and while I usually run it in text mode, it does have an X11 version that may pull in more dependencies than some people might prefer. I've also on occasion found tcptraceroute useful, and of course sometimes a Paris traceroute is good to have. Installing more than one program that has traceroute functionality in the default installation might be appropriate.
DESCRIPTION: Several places where the default behavior of vmbuilder could possibly be improved, relative to what seems to happen on 14.04 / 16.04:
I've found that I always end up wanting --addpkg acpid when running vmbuilder so that the host can send the guest a request to shut down cleanly; maybe include this package by default unless it is somehow explicitly deselected?
I have developed a habit of always using --addpkg linux-image-virtual because at one point I ran into problems when not using it; if it is still needed, it should probably be included by default.
I've ended up with VMs in a directory where I didn't intend to have them when not specifying the -d flag; perhaps it would be better if vmbuilder would refuse to run without a -d flag explicitly specifying the directory. (Trying to identify all of the options one needs the first time running vmbuilder can be overwhelming, leading to leaving some options out and then ending up with a suboptimal VM, and sometimes one doesn't want to start over and rebuild the VM with the correct options.)
It might also be desirable to make the --timezone option mandatory; I think the default behavior is to put the guest in GMT rather than having it inherit the host's timezone, which can be surprising, especially if the host's timezone had initially been autodetected by the installer.
FLAVOR: Ubuntu Desktop, Ubuntu Server, and Ubuntu Core
HEADLINE: open-vm-tools auto installation
DESCRIPTION: It would be nice if the installer would automatically determine whether it is running as a guest inside a hypervisor for which open-vm-tools is useful, and if so, automatically install open-vm-tools.
FLAVOR: Ubuntu Desktop, Ubuntu Server, and Ubuntu Core
HEADLINE: AS112 inspired mirror system
DESCRIPTION: https://www.as112.net/ describes a largely uncoordinated system for providing somewhat localized servers to handle certain DNS zones. It seems to me that something somewhat similar could work for anycasting mirrors of major free software distributions. I suspect that public peering point operators and ISPs might be most likely to participate if a single server could act as a mirror for Debian, Ubuntu, CentOS, etc. It would be best if the clients were set up to fetch a list of packages and their checksums from the centralized servers operated by the distribution maintainers, and then would try to fetch the packages from the local uncoordinated mirror, and if the local uncoordinated mirror either doesn't have the file or has the file with a bad checksum, would fall back to fetching the file from the official centralized server.
DESCRIPTION: 16.04 LTS shipped with a cacti package of a version written for PHP 5, but shipped PHP 7, and Ubuntu's effort to patch cacti for PHP 7 compatibility was incomplete. When I reported a bug with using more frequently than once a minute polling resulting from this, I got a response that seemed to indicate that Ubuntu was in no hurry to fix it. I ended up simply switching from Ubuntu to CentOS with the epel repository, which avoided both the bug I did report, and some other buggy behavior that I suspect may have a similar PHP version incompatibility root cause that I have not wasted the time to track down.
HEADLINE: DANE for TLS in Firefox, wget, curl, etc
DESCRIPTION: Support TLS server verification using TLSA DNS records protected by DNSSEC as described at http://www.internetsociety.org/articles/dane-taking-tls-auth... and https://en.wikipedia.org/wiki/DNS-based_Authentication_of_Na... ; this should have a smaller attack surface than the current mess of X.509 certificate authorities that are trusted by web browsers. Doing this well may require better client side DNSSEC validation; my impression is that DNSSEC validation deployments in the real world today often tend to have only the recursive resolver doing DNSSEC validation, with a potentially insecure connection between the client and the recursive resolver. Firefox probably ought to check the entire DNSSEC signature chain itself.
FLAVOR: Ubuntu Desktop, Ubuntu Server, and Ubuntu Core
HEADLINE: bounds checking gcc
DESCRIPTION: https://gcc.gnu.org/extensions.html mentions bounds checking patches for gcc. Get these patches updated to work correctly with the current version of gcc, and get most of the Ubuntu userland compiled with bounds checking enabled (and then gradually work on making more and more of the userland compatible with bounds checking, and also extend it to the kernel). I suspect paying for this development work would be cheaper than paying out a $10,000 bug bounty every time someone finds a bug that could have been rendered irrelevant by bounds checking support.
HEADLINE: bind package with support for DNS cookies
DESCRIPTION: https://kb.isc.org/article/AA-01387/0/DNS-Cookies-in-BIND-9.... describes DNS cookies; last I checked, it seemed that Ubuntu wasn't in any hurry to upgrade to a version of bind that turns DNS cookies on by default, and also probably wasn't passing the build time option to turn on DNS cookies on the version that was being shipped.
DESCRIPTION: When I install a new server that should have a static IP address on a network that has a DHCP server, it would be nice if the installer would give me the option to configure the static IP address, instead of it initially getting a DHCP lease and then needing to have the static IP address configured by editing /etc/network/interfaces after booting off the hard drive.
IIRC, Matt Yglesias and Paul Krugman have both been arguing that an inflation target of 4% to 5% instead of the current 2% or so, and perhaps also another round of Congress passing an economic stimulus package, would help a lot with the overall economic demand.
But I've never been convinced that there's really a good solution to the problems that Michael Moore's Roger and Me discussed, where good paying factory jobs get replaced with automation. pg has suggested that overpaying for labor was something that only worked for the Big Three while they were in startup mode, until foreign cars became popular. It would be fascinating to watch pg discuss this with Michael Moore, because Moore seems to feel entitled to something being sustained that pg seems to argue was clearly unsustainable. And I think I've seen an article that seemed to imply that the Tesla Factory hasn't employed very many former NUMMI workers because Tesla wants more skilled folks than NUMMI needed.
Maybe you want to convince Whole Foods to come to Santa Clara. (Although recent experience in Massachusetts tells me that there are benefits in terms of team member competence to going to a store that's been open for a while instead of a new store.)
Which of efficiency or price/watt matters more depends on the application.
For example, I believe something like 4 m^2, 40% efficient, 8 hours of charging is enough to get electricity to drive a Tesla something like 30-40 miles. (In practice, I assume you'd cover all of the outside of the car with 40% efficient solar panels to make this work (or maybe even 80% if you can figure out how to build them), and then 4 m^2 approximates the effective useful area after you take into account suboptimal sun angle, the side of the car facing away from the sun, etc.) The number of people whose commutes could be entirely covered by 40% efficient solar panels built into the car being used to recharge the battery is a lot more than the number of people whose commutes could be covered entirely by 12-15% efficient solar panels. Current mass produced technology doesn't make any variation on this cost effective today, but it seems clear that in the future, a 12% efficient panel array built into a Tesla is never going to be all that attractive for this application no matter how cheap it might become.
I suspect that high rise apartment buildings can cover a much larger percentage of their energy consumption from solar panels mounted on their sides and roof if those solar panels are 40-80% efficient instead of 12-15% efficient. If localized solar right next to the consumer of the power becomes cheaper than paying for electric grid transmission, this may become very relevant in 10 or 20 years.
Even for residential suburban solar installation, where finding adequate space for an adequate quantity 12% efficient solar panels is generally not a problem, the really important thing is not the price per watt of the panels themselves, but the price per watt of the system as a whole. The mechanical support of the panels is a part of the system cost, and the cost of that portion will shrink with more efficient solar panels, though of course there is an open question as to whether that will end up lowering the total system cost or not in the long run.
If those are 1,5000 professional tax preparers who each perpare taxes for many taxpayers, it might make sense.
Going after only high income citizens might also create a reasonable situation where $133k/taxpayer in their biased sample set might happen. It's often not unreasonable to focus only on those individuals who might produce the largest numbers...
If the production server is running as a VM to begin with, how hard is it to take a snapshot of the whole VM and replicate that? (It does sort of break down if you have a database that's continuously updated and you want to try to have a continuously updated copy of the database, though.)
Does the New York Times really want this sort of policy? Are they offering that anytime they write a story about someone, and that person doesn't like what the New York Times wrote, that the New York Times will be happy to simply remove such stories from their website?
On the other hand, if Tesla continues to raise funds every six months in a manner similar to the recent secondary offering, buying and holding the stock for a long time will probably not turn out to be the fantastic deal for shareholders that it has a good chance of being if Tesla chooses to bootstrap from here to 20 million cars/year.
Elon seems more interested in accelerating EV adoption than making his shareholders rich, so we'll probably see more of these secondary offerings if they can happen on terms Tesla finds reasonably good.
The dealers claim they are protecting consumers from the automakers.
I keep wondering where I can find an example of a person who clearly is not an owner/employee/etc of a dealer who genuinely feels protected. There are probably far more people who enjoy going to the Apple Store (which was created by the same guy who now works at Tesla to set up Tesla's showrooms) than who enjoy the dealership experience.
Also, in a lot of the rest of the world, there apparently are no independent dealers, and that seems to be pretty non-disasterous.
It might also be worth looking at where airplanes tanker fuel, that is to say, carry more fuel than they need for their current leg because refueling at the next stop would be difficult or expensive. Apparently a lot of that currently happens on short flights to small islands; while I hope Wright Electric and/or the EViation Alice will eventually take over that market, in the short term that's a market that might be willing to pay a bit more for liquid fuel made from air plus local solar panels.
Fuel oil that can be burned in combined cycle power plants in the winter may also be valuable for dealing with seasonal imbalances in demand vs renewable generation that lithium ion batteries can't cost effectively balance.