Article
Improving database record retrieval throughput with appengine Using AppEngine with Python and the HRD retrieving records sequentially (via an indexed field which is an incrementing integer timestamp) we ... (more)
Anyone know how to get more throughput from AppEngine database? We'll pay!
Several people have asked me what I am going to use instead of Google #Reader . I know what I definitely won't use - the majority of the alternatives out there, because they are obsessed with with showing items as rectangles with as much of the rectangle taken up with pictures as possible. (That is also why I gave up on G+ on Android.) My RSS feeds do not constitute a pretty magazine. For example look at http://feedly.com and notice how the screenshot of the mac is showing 6 articles - all that space for 6 articles! Pulse's home page has way more articles but all as images. 99% of my RSS articles have no images.
Reader has three important parts. One is the backend which means you can read from any number of computers and devices and a centralised location keeps track of your feeds and which articles have been read. There is no open standard protocol for this and I'm hoping that in the next few months one is born. (Reader was used by many but as far as I can tell the API was not official, arbitrary and reverse engineered.)
The second was updating the feeds (a background task). Having worked on consuming RSS feeds before, it turns out that many only have the most recent few articles which could turn out to be a few hours worth or at most a day. If you don't regularly poll the feeds then you will miss out on articles. This rules out pure clients like Liferea.
The final part is the presentation which worked well with a hierarchy of folders, feeds, articles and article, making it very easy to jump around hierarchy. UI that blends it all together into a single stream does not work (eg G+) because that only works with a low volume of articles. (I do not follow many people on G+ due to their posting volume. I would happily follow them via RSS feed but G+ doesn't export the data so I don't. Google loses by being a closed island.) So far only http://tt-rss.org shows that hierarchical context and navigation UI. Others like http://theoldreader.com/pages/tour seem obsessed with the whole social side, and note there is no screenshot showing the actual reading experience!
Oh, and the presentation part needs to work in disconnected mode on mobile too.
My plan is to punt for a few months hoping that someone comes up with the sweet spot of openness and readability. If not I'll just write my own. The database side of things can be solved in a home grown manner, but a toolkit like https://www.parse.com could be used instead to solve the data problem keeping web (Javascript) and mobile clients all synced up with state.
However I realised that it would be even easier to use +Dropbox as the state and synchronization mechanism. Each article becomes a file, and gets deleted by a reader when done. The poller just keeps adding the files from feeds. (Yes, that is the same principle as Maildir.) Dropbox released https://www.dropbox.com/developers/sync recently so that solves the offline mobile clients sync problem. I just wonder how well they will cope with hundreds of little files being created and deleted every hour.
Category: gplus – Tags: google reader, tt-rss
Google caused a DDOS attack against pretty much every other RSS #Reader out there. All the various sites are not responding, timing out, incompletely rendering etc. Like everyone else I'm concerned about how to replace one of the tools I spend most time in, especially from Google. (The first tab of my first window has been Reader for many years.)
Although RSS readers are conceptually simple, the UI is incredibly hard to get right and Reader managed to hit the right balance of simplicity and density without gratuitous slickness.
I'm actually quite happy not to #SaveGoogleReader - this is a sign and reminder not to be so dependent on Google.
Category: gplus – Tags: google reader
Article
Lab Tests Offered | Discount Blood Tests at the Lowest Prices Online Buy lab tests and discount blood tests online, privately and securely through Health One Labs online ... (more)
The first rule of the American healthcare industry is that under no circumstances do they say what anything costs, and especially not in advance. Generally they will flat out refuse to tell you beforehand or immediately afterwards, and any bills you get later will have a bewildering variety of numbers just to make sure you still don't know.
My health insurance is atypical in that it closely resembles actual insurance - I pay for everything unless something calamitous happens and after a several thousand dollar deductible, so I do care how much things cost. I needed some lab tests done and wanted to find out how much they would cost when I pay cash directly. It took talking to 5 different people on the phone at the lab service used by my doctor to finally get some numbers and they were outrageous. For example a hemoglobin/A1C test is $66 plus a $21 fee for the blood draw - you can get a kit at Walmart for $9 for the same thing. It also looks like they would have refused to do the tests unless my doctor faxed them an order, and the total would be huge.
Eventually I managed to discover Health One Labs. In an astonishing violation of the first rule they put the prices right up there on the website. The value packages are well priced, most being under $100 with individual tests around 20% of the price the normal lab the doctor uses.
I tried it out. Four minutes after entering my credit card information, I received a PDF with a lab order on it (and no need to get the doctor involved). The actual blood draw is done at a company named Lab Corp - they have 13 locations within 25 miles of me - no extra fees or charges, or appointments and was seen immediately. 24 hours later (almost to the minute) I received my results.
This should be the normal experience, but is so remarkable in the US healthcare system that I had to write about it.
Category: gplus
I've always been curious why there are so many +Android backup applications. Technically the backups are trivial - apks are at a known place and each app's data is at a known place, so just grab that stuff for backups and put back for restores. Android 4 has adb backup/restore commands which do pretty much that. Simple, effective, what could go wrong? Short answer: it is quicker to just delete your data. Those backup apps exist for a reason.
Yesterday I decided to wipe my phone and tablet back to factory images. In particular I wanted to encrypt them, and I did want root, but didn't need +ClockworkMod recoveries which rooting usually require you install first. +Chainfire has a CF-AutoRoot that does that for you.
Backing up the phone gave me a ~400MB archive (I didn't include apks since they can be easily redownloaded). I did the wipe/factory/root/encrypt dance and then did my restore. It had no effect as far as I can tell. And I had to manually select every app I wanted again in the Play Store. Doing another restore after that still had no effect.
Having learned my lesson, I did a backup of the tablet including all apks. It was ~12GB. Did the wipe/factory/root/encrypt dance and then did the restore. It restores one apk and then the Android gui crashes, restarting itself a few seconds later. This is completely repeatable. Gui crashes don't have a crash reporter so Google won't even know this happening. And I'd love to hear the explanation as to why a restore would crash the whole gui environment.
I could of course manually dig things out of the backups, but that would be far too time consuming. There is no unique data on the devices so I didn't actually lose anything. But the reinstall of every app, having to supply usernames and passwords to so many of them, changing settings, and placing them in the right places in the launcher is really tedious.
Article
Dreamliner woes: Are sweaty passengers to blame? – Is it possible that one of the key benefits of the Boeing 787 Dreamliner, which is denser and more humid ... (more)
Interesting speculation about a potential cause/correlator with the 787 battery issues - hundreds of breathing perspiring passengers create very different environmental conditions, and especially different than during certification.
Nicotine would get everywhere and had the advantage of being visible back in the days when smoking was allowed. (See also China Airlines flight 611.)
I highly recommend this video showing how blind users use screen readers. You'll be surprised at just how fast they run, as well as ways they distinguish case.
Category: gplus – Tags: recommendation
Fun things to do to the human brain. His "ad" for a cruise ship is also excellent (NSFW) F***IN' CRUISE SHIP!
(Via Reddit)
+Andy Glew speculates about Google news favouring stories about driverless cars. All it takes is watching one of the car crash compilations on youtube to make you wonder why we even let people drive cars.
One interesting aspect to me is that there are situations where the laws of physics mean an accident is definitely going to happen - the kinetic energy has to go somewhere when you are less than the stopping distance away from other vehicles. What should the driverless cars do in that situation? For example it could try to keep the minimum number of vehicles involved, or the minimum number of people. Is it better to have fewer people hurt worse than more people hurt lighter? Full width of the vehicle collisions are better than partial width because there is more in the way to absorb the energy. Given the choice between hitting two vehicles should the one with better safety features/rating be the target?
Will cars start having something like TCAS as they do in aviation? When multiple cars have to move out the way to prevent an accident it provides a negotiation model. Will cars with drivers also have it?
It will become in everyone's interest to have cameras as black boxes (pretty much standard in Russia due to insurance fraud which is why so many videos come from there). In any crash where a driverless car is involved that car will need them to prove it wasn't at fault. And of course the 90% of drivers who think they are above average will want them to prove the driverless cars or other drivers were at fault.
Category: gplus – Tags: driverless cars
I've used the Google Chrome To Phone extension for two years now. In theory it is very simple - while on a page (or selected text) in desktop Chrome, press the button and the page/text show up instantly on your Android device. Behind the scenes it is using c2dm gcm and serves as a useful demo including source code.
In practise it isn't as good. The "instant" is sometimes instant, and sometimes never, and sometimes a random time period inbetween. Rebooting a device can result in over a month's worth of content all showing up at once. I also have more than one Android device (perils of being a developer) and in theory the url should be sent to all of them. Again it is very unreliable.
Today I finally got rid of the extension from Chrome and the app from my devices, switching to Pocket based on the recommendation of +Stephane Verdy . Looking good so far ...