The Daily Parker

I just released a couple of minor fixes to Weather Now. Build 8126 has slightly tidier top and nav bars, and I can configure the front page "latest weather" list on the fly. Previously the list lived in a static application config file that I could only change by redeploying the app.

Enjoy.

At the end of the month, I'm taking the first real vacation I've had since 2017, to Central Europe. After connecting through Heathrow, I land in Prague, Czechia; then by train on to Vienna, Austria; then Salzburg, Austria; then a flight back to Gatwick and a night in London. And because of Vienna's and Salzburg's proximity to Austria's borders, I will probably also visit Slovakia, Hungary, and Germany—at least for a few minutes.

To prepare for this trip, about a month ago I downloaded Duolingo, and started the Czech program. I also jumped into the German program at unit 5, as I've studied German before.

I've had mixed results.

First, I want to make it clear that I love Duolingo. I have learned some basic Czech and I've gotten my German back to tourist-level fluency. When I get back to the US, I'm planning to load up French and Spanish, with the goal of getting both back to conversational levels. Just practicing languages every day keeps me learning them, so I believe I'll eventually finish the French and Spanish programs with some pretty good skills in both.

As for my upcoming trip, I've decided to change my approach. Thus far, I've spent about 20 minutes a day on Czech and 10 on German. And yet I'm going through the German lessons much faster, for a number of reasons, not least of which is that I first learned German in high school and I first learned Czech 37 days ago.

In the German program, I'm breezing through things like „wie ist das Wetter in Wien?” and „entschuldigung, wo ist der Geldautomat”, both of which which I actually want to know, and I'm acing (almost) all the speech and listening exercises. (Im and in gave me a bit of a bother for a hot minute.) I've gone from my start in section 1, unit 5 to section 2, unit 4, and the app says I've learned about 300 new German words.

In the Czech program, by contrast, 37 days have gotten me to...section 1, unit 5. And that's only because I gave up on the optional grammar drills after unit 4. I can say things like „jsem David” (I'm David) and „jsou to zvláštní zvířata” ("those are strange animals"), but not every time, and with no guarantee of grammatical accuracy. You see, Czech is a declined language, where all the grammar lives at the ends of words. I just can't seem to get the correct word endings 6 times out of 10. It's supremely frustrating.

So starting today, I'm going to change my approach.

First, I'm going to flip my priorities and spend 2/3 or more of my time on German. That's closer to my trip plan, anyway: from wheels down at Václav Havel Airport to my train crossing the Austrian border, I'll spend at most 48 hours of the 7-day trip in Czechia.

Second, I'm going to concentrate on Czech vocabulary, not mastery. For example, I'm going to skip the grammar drills at the end of each Czech unit and concentrate on just getting enough sentences right to move on to the next unit.

I'll continue to do the German drills, though. This will be my 5th trip to German-speaking countries, and will not be my last, but I have no idea if I'll ever get back to Czechia after this month. I'm singing Bruckner next year and probably Bach in 2025, but I have never to my knowledge sung in Czech. And I'm far more likely to remember the difference between „wo ist der Bahnhof” and „wo ist die U-Bahnstation” than I am to recall (or even say) „Jsem velký kluk” vs. „Jsme velcí kluci”.

I only hope „jsi hezká” comes in handy at least once...

I'm chasing down a bug that caused what we in the biz call "unexpected results" and the end-users call "wrong." I've fixed it in both our API and our UI, but in order to test it, I need the API built in our dev/test environment. That takes about 18 minutes. Plenty of time to read all of this:

• Ruth Marcus wants you to really think about Jean Carroll's testimony in her defamation suit against the XPOTUS.
• Tom Nichols sees narcissism, not "misguided patriotism," as the root of the classified intelligence leaks endangering the US.
• Jamelle Boule reminds the Chief Justice of the United States that when Congress calls, he needs to answer.
• Alex Shephard tells us what we can all see: House Speaker Kevin McCarthy (R-CA) sucks at his job.
• Tim Murphy can't wait for Twitter to die so news stories no longer quote "another Twitter user" ever again.
• ProPublica exposes the dangers children face when trains park across their route to school, forcing them to crawl over (or under!) them.

Finally, the Times explains how last year's 257 traffic fatalities in New York City undermine the claims that "Vision Zero" is working. But Strong Towns already told you that.

OK, build succeeded, fix is now in Dev/Test...on with the show!

As professional narcissist Elon Musk threatened, on Thursday Twitter abruptly ended their verified "blue check" program. Suddenly, Twitter users had no way to know for sure whether tens of thousands of government agencies, celebrities, journalists, and other people whose jobs depend to some extent on their credibility, were who their Twitter accounts purported to be.

It only took two days for someone to hoax the City of Chicago:

Impostors posing as Chicago government officials, including Mayor Lori Lightfoot, posted a series of tweets early Friday morning falsely claiming that North Jean Baptiste Point DuSable Lake Shore Drive is being permanently closed.

The accounts posting the false tweets claimed to be Lightfoot, the Chicago Department of Transportation and the Illinois Department of Transportation. Their posts were seen by well over 100,000 people early in the day, according to Twitter data.

The impersonation effort came just a day after Twitter removed verification from accounts that had been previously verified for the purpose of credibility. Twitter CEO Elon Musk’s controversial decision to remove the “legacy” verifications took away the authenticated status of all three of the impersonated accounts.

Before Friday, the usernames of the true mayoral account, CDOT account and IDOT account had all previously appeared alongside a blue check mark signaling that the accounts had been proved to actually represent the authorities they claimed to represent. However, those check marks were taken away Thursday because of Musk’s decision.

New York City's government also had their own problem.

But this highlights a real problem: in a disaster, or an election, how will people know what information is real? I think "by not using Twitter" seems like the right answer but I also don't think Twitter will fully die by next November.

I didn't publish on Twitter very much before. Today, I'm just standing on shore, watching the ship sink. But it's a big ship, and its sinking will foul the environment for a long time.

First, because it's April 20th, we have a a couple of stories about marijuana:

• Daniel Wolfe calls on cannabis sellers to pick names that actually provide brand differentiation.
• Cannabis companies here in Chicago (and, one expects, elsewhere) provide buses to private events to get around public-consumption laws.

(The Daily Parker owns shares in Chicago-based Green Thumb Industries.)

Second, because it's the 21st Century, we have a collection of articles about the end of democracy:

• Thomas Edsall warns that the Right has adopted a strategy of abusing the institutions of our democracy to undermine it.
• Elon Musk, whose 'sploding rocket this morning should underscore that he knows about as much rocket science as he does software, wants to become a latter-day Pullman.
• Dunning-Kruger poster child Mike Lindell must pay a computer engineer $5m after the engineer proved conclusively that Lindell's "stolen election" data was bogus.
• Josh Marshall looks at Florida governor Ron DeSantis' (R) trip to Washington that managed to get several sitting Representatives (including one from Florida!) to endorse the XPOTUS. As Marshall says, "Stick a spork in him. He's done."

And in me. I've got software to write.

The clouds have moved off to the east, so it's a bit warmer and a lot sunnier than yesterday. I still have to wait for an automated build to run. For some reason (which I will have to track down after lunch), our CI builds have gone from 22 minutes to 37. Somewhere in the 90 kB of logs I'll find out why.

Meanwhile, happy Fox News On Trial Day:

• Jennifer Rubin foresees years of aftershocks from the Tennessee legislature's expulsion of two Black members last week.
• Why are right-wingers making up conspiracy theories about vegetarian food, now?
• Jaime Green re-reads Carl Sagan's novel Contact.
• WeWork has closed its four-floor Chicago office (one of several in the city) while facing a claim in Cook County Circuit Court for $360,000 in back rent.
• Vox's culture reporter Aja Romano muses about how the third season of The Mandalorian has given the Star Wars franchise "a Baby Yoda problem."
• The New Yorker interviews Luis von Ahn, creator of the Duolingo language-teaching app I'm using to prepare for an upcoming trip to Czechia.

Finally, I've started reading The Odyssey, so I applaud National Geographic's article this month on the history of the ancient world in which Homer set the poem.

Not five minutes after my last post, I discovered a completely borked feature, caused by a change to the way Azure.Data.Tables executes queries.

The Daily Temperatures feature stores data in the same table as the History feature. Each row represents a weather report, where the table partition key is the weather station identifier and the row key is the date and time of the report. So, for example, the first row of data for Chicago-O'Hare in the 2023 table has a partition key of KORD and a row key of 20230101-0051.

Climate records use a row key of "Climate-" and the date. So yesterday's climate data for Chicago-O'Hare has a partition key of KORD and a row key of "Climate-20230415". Easy to remember, and easy to construct queries.

To that end, the original (.NET 6) code looked like this:

var query = 
	from entity in table.CreateQuery<ClimateRecordTableServiceEntity>()
	where entity.PartitionKey == locationId
		&& string.Compare(entity.RowKey, lowerRowKey, StringComparison.InvariantCultureIgnoreCase) >= 0
		&& string.Compare(entity.RowKey, upperRowKey, StringComparison.InvariantCultureIgnoreCase) <= 0
	select entity;

When I upgraded to .NET 7, I naïvely just changed the first line, to this:

var query = 
	from entity in table.Query<ClimateRecordTableServiceEntity>()
	where entity.PartitionKey == locationId
		&& string.Compare(entity.RowKey, lowerRowKey, StringComparison.InvariantCultureIgnoreCase) >= 0
		&& string.Compare(entity.RowKey, upperRowKey, StringComparison.InvariantCultureIgnoreCase) <= 0
	select entity;

When confronted with a 30-day query, though, it spun off into the abyss and crashed the whole app.

The correct code looks like this:

var query = table.QueryAsync<ClimateRecordTableServiceEntity>(entity =>
	entity.PartitionKey == locationId
		&& string.Compare(entity.RowKey, lowerRowKey, StringComparison.InvariantCultureIgnoreCase) >= 0
		&& string.Compare(entity.RowKey, upperRowKey, StringComparison.InvariantCultureIgnoreCase) <= 0);

See, now the filter part of the query goes inside the method call. (There's an extra step in reading the async results back, too.)

So the effect of the naïve fix was to hit the table 30 times getting back the entire partition each time. Remember that all of the weather reports go into the table? So, yeah, the 2023 table already has something like 7.5 million rows, or about 2,500 in each partition. So it tried to read 75,000 rows just to bring back 30. Oopsi.

I'm deploying the fix now.

It took a few weeks at odd hours, but I have finally deployed the latest version of Weather Now (5.0.8507). I didn't update anything visual, but all the plumbing got a refresh. It's now running in .NET 7 (until November, when .NET 8 comes out), and I did a top-to-bottom review of its asynchronous code.

The app now runs noticeably faster, and I believe the corrections to the async bits will cure the nagging (but invisible) problem of thread exhaustion that happened from time to time.

Now I can start making some other fixes and adding some long-missing features. Really, I just needed to get back into the swing of it.

I spent all day updating my real job's software to .NET 7, and to predominantly asynchronous operation throughout. Now I have four stubbornly failing unit tests that lead me to suspect I got something wrong in the async timing somewhere. It's four out of 507, so most of today's work went fine.

Meanwhile, the following stories have backed up:

• The Economist wonders what our friends should think of the XPOTUS's dog and pony show.
• Alex Shephard worries that television and cable news just don't have the skill to cover the 2024 election, after seeing their "dumb, desperate" coverage of the XPOTUS's indictment.
• Justice Clarence Thomas (R) has rich Republican friends who took him and his ethically-challenged wife on luxury vacations, but never disclosed them as required by law, adding to his legendary status as poster child for term limits.
• Paul Krugman can't figure out why Republican voters keep electing people who want them dead.
• Illinois had a cool, wet start to spring, but we're about to have a heat wave, with temperatures predicted to hit 11°C (20°F) above normal next week.

Finally, a very rich person is very annoyed after his or her private jet got stuck in the mud at Aspen's airport. It seems the guy sent to pull it out of the mud maybe needed another lesson on how planes work, because he managed to snap the nose gear right off the $3.5 million airplane. Oopsi. (There's video!)

I woke up this morning to about 450 error messages because our team's Twitter account got suspended at midnight UTC—that is, at 7pm last evening. No one knew about it because we never considered Twitter errors critical enough to keep them in our inboxes; they all go to an Outlook subfolder. Apparently, Twitter finally decided that our 15-minutes-apart API calls violated a policy, but we never got informed that this would happen or that we needed to correct something.

As near as I can figure out, Twitter's new pricing structure gives developers access to post 1,500 Tweets per month for free, but doesn't allow searches. Our app doesn't post, it only reads Tweets. The "Basic" level costs $100 and allows reading 10,000 Tweets per month. Anything more than that and you have to apply for an Enterprise license—and they don't disclose pricing information online.

Our app read about 12,000 Tweets per hour because each API call brought back about 2,000 Tweets at a time. And as far as we knew, that was what we signed up for. We even have code that checks whether we're approaching the API rate limit so we can pause the API calls.

All of this happened on the same day that Twitter decided National Public Radio is "state-affiliated media," i.e., in the same category as TASS and North Korea's "news" channel. NPR is not amused:

NPR operates independently of the U.S. government. And while federal money is important to the overall public media system, NPR gets less than 1% of its annual budget, on average, from federal sources.

Noting the millions of listeners who support and rely upon NPR for "independent, fact-based journalism," NPR CEO John Lansing stated, "NPR stands for freedom of speech and holding the powerful accountable. It is unacceptable for Twitter to label us this way. A vigorous, vibrant free press is essential to the health of our democracy."

NPR officials have asked Twitter to remove the label. They initially assumed it was applied by mistake, NPR spokesperson Isabel Lara said. "We were not warned. It happened quite suddenly last night," Lara said.

In response to an NPR email for this story seeking comment and requesting details about what in particular might have led to the new designation, the company's press account auto-replied with a poop emoji — a message it has been sending to journalists for weeks.

Mastadon user Rod Hilton posted this in December, which perfectly captures  the value of Twitter's infantile owner:

He talked about electric cars. I don't know anything about cars, so when people said he was a genius I figured he must be a genius.

Then he talked about rockets. I don't know anything about rockets, so when people said he was a genius I figured he must be a genius.

Now he talks about software. I happen to know a lot about software & Elon Musk is saying the stupidest shit I've ever heard anyone say, so when people say he's a genius I figure I should stay the hell away from his cars and rockets.

I figure, I'll keep Twitter as long as some of the people I like keep posting on it, but I know the app will eventually fail. It's a little annoying that our research at work has to stop, because now I have to build an API adapter for a new app.

I can't help but compare Musk to Eddie Lampert, the guy who destroyed the department store Sears. I despise sociopaths like Lampert, but at least Lampert had a definable business strategy and extracted value from tearing the brand apart. Musk really isn't all that smart, and Twitter isn't all that valuable. "Say what you want about the tenets of National Socialism, Dude. At least it's an ethos."