A Concurrent Affair

I’ve been exclusively working from home now for over a year. The last day I was at the office was Friday, February 28, 2020. At that point, I decided the risk of getting on public transportation and being around hundreds of co-workers was too high.

Early during the next week, we stocked up and beat the rush on supplies and toilet paper, which allowed us to not shop for several weeks. In retrospect, we were probably a little too concerned (for example, we let mail sit for a couple of days before opening it, even though fomite transmission has later shown to not be a problem), but we just didn’t know. And it was relaxing to not need to go anyway.

I had a pretty good setup at home already, with a standing desk and two vertical 20″ screens (which I’ve had since early in grad school). I got the standing desk because standing was the only way I my infant daughter would let me work. The standing desk definitely saved my back. Early on, I also used a $300 subsidy from my employer Oracle to buy a 27″ 4K monitor. The screen real estate without having things split between the 20″ monitors definitely helps. I just wish I had got a 32″ 4K monitor. On the 27″, the pixels are too small.

Our schedule has definitely changed. My commute from home to work was actually pretty good, but I nonetheless save about 2 hours combined. This has allowed us to actually work out every morning. Again, if we hadn’t started that in March 2020, this past year probably would have been much less healthy for us. I’ve read that about 60% of Americans have had undesired weight changes (mostly gains) during the last year. Our weight has stayed the same, but we’ve definitely gained muscle mass. I don’t think I’ve ever been this fit on a whole-body level. My legs are less fit than during climbing season, particularly the 2017 Rainier season, but overall, my body is doing great.

I probably still spend more time working than I did before. Especially in the beginning, in March and April 2020, it was difficult for me to set boundaries. I worked early, I worked late, I worked a lot. That was draining, and I had to find the self-discipline to not look at the work computer even though it is right there. In general, self-discipline has become much more important.

Initially, I thought I’d quarantine and work from home for maybe a month or two. Now we’re reaching the end of month 13, and it will still take a few more month for sure, even though an end is in sight.

Microsoft recently announced that they are letting employees back on campus, even though they are still encouraging employees to work from home. Oracle hasn’t made that announcement yet, and I don’t think anyone will be required to come back at least until they are vaccinated, but Microsoft’s announcement made me search my feelings about returning to the office.

I’ve realized I’m not close to ready.

I anticipate it will be a few more months, maybe three, of working from home exclusively. And even after that, I plan to work from home most days of the week. There simply are lots of benefits, and my team and I have demonstrated that we can be productive even though we work remotely.

I listened to Moshe Vardi’s World Logic Day talk “From Aristotle to the iPhone” (recording to be released later). The chain of influences was quite illuminating.

That made me think of how I fit into the world of mathematics. I know I had calculated my Erdős number at one point, but I couldn’t find it anymore. I myself haven’t authored any mathematics papers, of course, but there are computer scientists who bridge that gap.

So I used a list of Erdős collaborators, which I confirmed using the MathSciNet tool. Then I used a similar tool for computer scientists from csauthors.net.

It turns out my Erdős numbers is at most 4 (I didn’t check all of Erdős’ collaborators, so it could be lower, although I find that unlikely).

Here’s the collaboration chain:

4. Mathias Ricken co-authored 2 papers with
3. Vivek Sarkar co-authored 2 papers with
2. Nimrod Megiddo co-authored 3 papers with
1. Miklós Ajtai co-authored 1 paper with
0. Paul Erdős

Small, small world.

Update:

Turns out I didn’t have to do any of this search manually. My author page on csauthors.net already states:

Collaborative distances:

Dijkstra number of four.
Erdős number of four.

Over the holidays, I migrated from 1&1 to DreamHost, and I got almost everything to work. The only thing that didn’t work anymore were the email notifications from my Buffalo LinkStation NAS. Normally, they report once a day that everything is okay, or immediately if there is an error.

I had re-created the email account that I use for that in DreamHost, updated the SMTP server and credentials in the LinkStation admin interface, but it wasn’t sending any emails. It was saying “Test email sent.” when I tested the new settings, but nothing ever arrived. No combination of login mode and SSL/TLS/disable worked.

I had to figure out if there were logs anywhere. I know the NAS has SSH access, but I hadn’t done that in a really long time. It seems like somehow, it got disabled again. No problem, yesterday evening I ran an easy-to-use Java application called ACP Commander GUI to re-enable SSH access and set a new root password. Then I could SSH into the NAS.

The logs were fortunately easy to find in /var/log. The file messages contained the following line every time I sent a test message:

550 5.7.1 Sender domain not allowed

What in the world does that mean? It turns out DreamHost has a policy against spoofing the sender. But I don’t think I was doing that. All I was doing was sending it from nas.machine@domain.com (which had the email account set up with DreamHost) to personal.email@gmail.com. I wanted the sender to be nas.machine@domain.com, which shouldn’t be blocked, and I wasn’t setting a different sender email anywhere!

After searching a bit for the ssmtp.conf file, I found it at /etc/melco/ssmtp.conf, and it contained the following line:

FromLineOverride=YES

Whenever I set it to NO, it got overwritten again and set to YES. I don’t know from where or why, but I assume this somehow sets the sender email to the destination email. So when I was trying to send to personal.email@gmail.com, the NAS was inadvertently also setting the sender to personal.email@gmail.com, which was prohibited by the DreamHost policy.

This means I can’t send to personal.email@gmail.com from the NAS. Instead, I have to send to an email account on my DreamHost domain (e.g. nas.forward@domain.com) and forward it from there.

Oh, I love troubleshooting my tech in my free time.

I moved the blog from 1and1 to DreamHost. Please let me know if you encounter any problems.

A little while ago, we installed a Ring alarm system in our house. We like it quite a bit. The initial installation hurdle was the need for three range extenders; then, two motion detectors were unreliable, but that turned out to be a battery issue. For the last month or so, it has run without a hitch.

Price was one of the main reasons to go with Ring instead of SimpliSafe. Both the equipment and the professional monitoring was cheaper (monitoring is $10/month or $100/year for Ring, compared to at least $15/month with SimpliSafe). Compared to ADT and Brinks, we didn’t want a long-term contract.

One thing that had me pause for a few days before purchasing the Ring system was the lack of a glass break detector. It’s really surprising that Ring doesn’t have one. In the end, though, I convinced myself that it’s really not a big issue. There are two situations: either we’re home and the system is armed, or we’re not home and the system is armed. If we’re not home, then the motion detectors are armed, and anyone stepping into our house will be detected. If we are home, then the system is armed in such a way that the motion detectors are disabled, but I’m pretty sure we’d hear the sound of glass breaking.

Still, this bugged me a little. I ended up buying another Ring contact sensor, the same little sensors we’re using on all our doors and windows. I opened it up, and it has a little magnetic reed switch in it. If I don’t use the magnet, and instead use a couple of wires to bypass the reed switch, I can simulate opening and closing the contact switch by disconnecting or connecting the wires, respectively.

There are third-party glass break detectors, for example the Honeywell IntelliSense FG-1625. It has a “normally closed” connection, which is exactly what I need: Normally, I want the wires connected, thereby bypassing the (always open) reed switch. If breaking glass is detected, I want the wires to be disconnected.

I wired up both the Ring contact sensor and the Honeywell glass break sensor accordingly, but unfortunately, the Ring wouldn’t register the glass break sensor opening the connection. In the end, I added a 2k Ohm resistor as an end-of-line resistor from C (“closed”) to EOL (“end-of-line”), and now it appears to work. When nothing is happening, Ring shows the contact sensor as closed. When the glass break sensor detects breaking glass, Ring shows the contact sensor as open, and I can alarm on it.

I’m trying this out, for now, without actually alarming and possibly calling the cops over. I’m looking for two things:

1. Are there false positives? I’ve got the combined sensor sitting in our living room, where the most action happens. But unless we actually break glass, I shouldn’t see an alarm.
2. What’s the battery life for the glass break sensor? It’s not really meant to be run on a 9 V battery. If the battery drains too quickly (which would register as “open connection” and therefore “glass break”, unfortunately), I probably have to hard-wire it. See update below.

Update: A day in, and the battery has drained sufficiently that it’s not outputting 9 V anymore, but only about 6.4 V. That was enough for the glass break detector to go offline (and show as “open” in Ring). I’ve now hardwired it to a 9 V power supply, and am continuing to test item #1 above.

On Friday, I went on my first solo hike this year. It’s truly a different year.

I had never been in the Mowich Lake area of Mount Rainier National Park, and it looked like there were a number of good-looking peaks to visit. It had rained the days before, but I had the day off. I left home around 4:30 AM and arrived at Mowich Lake Campground right around sunrise, just as planned.

From the Mowich Lake Campground, I headed north on the east shore past the patrol cabin. Then I followed a boot path up in the direction of Knapsack Pass. At about 5400 feet elevation, I turned south to follow the Fay Peak trail. Finding this turn-off was a bit tricky. The map seemed to indicate I should climb up a small waterfall, but upon reversing my track, I found a short westward section that crossed a small creek and then immediately turned south towards the Fay Peak ridge.

I continued following that trail all the way onto to the ridge. I thought about climbing the rocks at the west end of the ridge, where the boot path arrives, but it was too slippery. In general, I was soaked already from squeezing past trees, and I put my rain pants and rain jacket on. This way, at least I was wet and warm instead of wet and cold.

On the ridge, I traversed the boot path towards Fay Peak. Mount Rainier showed itself briefly through the low clouds for the first time.

Initially, I attempted to scramble it from the west side, but with the slippery rocks, that wasn’t all that easy. I succeeded on the northwestern flank, but then discovered that it would have been a walk-up from the east (not too surprising, looking at the map). My next objectives, East Fay Peak and First Mother Mountain, still lay in the mist.

I really wasn’t sure I would do much more, it was so wet and slippery, and I didn’t want to do any dangerous scrambling. But I could see the trail down from Knapsack Pass, and decided it was probably easier to continue along Fay Peak ridge to Knapsack Pass. But the other peaks were optional now. Well, they all are always optional.

I continued along the ridge past the middle rocks, which I climbed, to East Fay Peak. Judging by the map, it looks like it would be easier to climb from the south, and I got almost to the top, but again, the last few feet were very slippery. I circled around on the north side of the peak, and again, found the peak was a walk-up from the east.

The weather was starting to clear, and First Mother Mountain looked amazingly beautiful. I sloped north and down to the path up to Knapsack Pass, to which I connected just below the pass. I decided to head up to the pass, take a break, and look at First Mother for a bit.

The closer I got to First Mother Mountain, the more awesome it looked, but it also got scarier, and I was getting tired. I decided to take a look and only scramble it if it’s also a walk-up, which it was.

The trail goes up to the westernmost little peak of the Mother Mountain ridge, then behind the 2nd little peak on the north side. First Mother Mountain can easily be scrambled from the northwest side.

In the picture below, taken on the summit of First Mother Mountain, you can pretty much see the entire traverse I did. I came up on the far side of Fay Peak, the big massif in the center background. Then I traversed the ridge to the left, with East Fay Peak at the far left. The path up to First Mother Mountain goes up the heather slope in the center, around and behind the rock tower, and then up the chossy slope in the foreground.

After that, I went back down to Knapsack Pass and down the main trail to the lake and the campground. It was a wet suffer fest — almost every step, I could wring out my gloves — but so much fun. The whole day there were only goats and I.

Do you guys want to know how to refer to the /instances node under paths using a JSON pointer?

#/paths/~1instances

~1 is the escape sequence for /. And ~0 is the escape sequence for ~. I’m really learning more than I want to know.

Why ~ and not \, for example? Okay, maybe you don’t want to have a third escape sequence, \\ to denote \. But then, why not ~~ for ~ and ~/ for /?

I guarantee you, I will never be able to remember if ~1 is slash or tilde.

<~1rant>

We got married! It was exactly the wedding we wanted to have.

I would really like a Lombok @Builder annotation that properly deals with required parameters at compile time.

Here is what I have in mind:

This should generate the following:

So basically, if we have n required parameters, we 2^n builder classes. For each r required parameters that have been set already, we need n-choose-r builders.

In the example above, we have three required parameters. We have one builder that has no required parameters set, three that have one value set, three that have two values set, and one that has all of them set.

Unfortunately, the source code for rewriting the bytecode in response to a @Builder annotation is almost 1000 lines long. Not exactly the best example of well-structured code.

Now that Amazon Music is shutting down the ability to upload and play your own songs, what do you use as a replacement?

I have hundreds of CDs that I imported — complete Mike Oldfield, almost complete Depeche Mode, lots of classical music of which I’ve found one performance I prefer.

How will I get those to play on my Amazon Echo devices and my Android phone?

We’ve been shopping around for a handheld GPS, since many climbs require the ability to retrace our route. Since I have an InReach SE, it made sense to check out the InReach Explorer+, which combines a satellite messenger with topographic maps. We also tried out the GPSMAP 64st. Both devices are by Garmin.

We took both up on our August climb of Mount Rainier. The InReach Explorer+ is nice, since it means we only have to carry one device, instead of a messenger and something with topographic maps. Unfortunately, in terms of map detail and accuracy, it is quite far behind the GPSMAP 64st.

Here are our tracks, from just before the Disappointment Cleaver to somewhere on the Emmons or Winthrop glaciers, as seen in Google Earth. First, the track recorded by the GPSMAP 64st:

Track recorded by GPSMAP 64st

And here is the track recorded by the InReach Explorer+:

Track recorded by InReach Explorer+

In terms of detail that can be discerned on the display, the InReach also can’t compete with the GPSMAP 64st. Here is the topographic map around Camp Muir, first on the GPSMAP 64st:

Topographic map on GPSMAP 64st

And here is the same map location on the InReach Explorer+:

Topographic map on InReach Explorer+

You can’t even see Muir Peak, to the east of the camp!

I think we’ll return the InReach Explorer+ and keep the GPSMAP 64st. We just have to figure out a way to keep the latter from turning on accidentally and draining the battery.

I’m using Kaspersky SafeMoney to redirect requests to websites that deserve a bit more security to a sandboxed browser. I don’t have any extensions in it, except for LastPass. I think it makes sense to still get the benefits of a strong master password and different strong passwords for each site that I don’t have to remember.

Unfortunately, when I added Yubikey to the mix, LastPass started asking me for the Yubikey every time the secure browser launches.

The option to “trust this browser for 30 days” doesn’t seem to do anything anymore. Has anyone got this to work?

I’d like to say I began to adopt two-factor authentication pretty early. I’ve had it enabled for Google for a long time, I jumped on it when Amazon and Dropbox made it available, I enjoy it on several of my banking accounts, of course.

At work, I got a Yubikey a few months ago, a hardware token that is used in combination with a password. It’s the “have” part of the “something you know, something you have, and something you are” philosophy of protecting secrets. A fingerprint or iris scan would be “something you are.”

With all of the recent hacks and ransomware attacks, I decided to beef up security at home as well. I ordered two Yubikeys, a primary one and one to lock away as a backup. As primary Yubikey I chose the Yubikey NEO, which has NFC and can be used with a phone. As backup, I chose the cheap Yubikey FIDO U2F key.

Unfortunately, the results are a mixed bag.

Yubikey for LastPass requires LastPass Premium ($12 a year, plus taxes). I signed up, then found out that LastPass Premium does not seem to support the cheap Yubikey FIDO U2F key. So I couldn’t use that one as a backup for LastPass.

KeePass on Windows has a plug-in that allows a Yubikey to provide one-time pads (OTPs). Again, this doesn’t work on the cheapest key. Also, it’s a bit cumbersome, since it requires at least three presses of the Yubikey button, for three separate OTPs. Initially, I had also configured this for 2nd slot on the Yubikey, which meant I had to hold the button for three seconds, three times (the 2nd slot is triggered by a 3-second press; the first slot is triggered by a simple press of the button).

I also used KeePassDroid on my phone, and that app plain and simply does not support OTPs at all. I could have programmed the NEO for a static password, though.

There is a newer app, KeePass2Android, that does apparently support getting OTPs from a Yubikey NEO, but I haven’t tried it yet.

LastPass on Android does support the NEO, but with NFC, you cannot switch which slot is going to be used. By default, it’s always slot 1. That means I can’t use NFC for both LastPass and KeePass2Android, because both would be using the same slot, and that doesn’t work.

So, I have hardware two-factor authentication in a bunch of places now, but I’m not sure I’ve added that much security. I strengthened some of my most important passwords, though. Maybe that “something you know” part is still most important after all.