Probably best to get this out of the way before this weekend:

If I meet you at a technical conference, you’ll probably see me extend my elbow in your direction, rather than my hand. This is because I won’t shake your hand at a conference.

People sometimes joke about “con crud”, but the amount of lost productivity and human misery generated by conference-transmitted sickness is not funny. Personally, by the time the year is out, I will most likely have attended 5 conferences. This means that if I get sick at each one, I will spend more than a month out of the year out of commission being sick.

When I tell people this, they think I’m a germophobe. But, in all likelihood, I won’t be the one getting sick. I already have 10 years of building up herd immunity to the set of minor ailments that afflict the international Python-conference-attending community. It’s true that I don’t particularly want to get sick myself, but I happily shake people’s hands in more moderately-sized social gatherings. I’ve had a cold before and I’ve had one again; I have no illusion that ritually dousing myself in Purell every day will make me immune to all disease.

I’m not shaking your hand because I don’t want you to get sick. Please don’t be weird about it!

Remember that thing I said in my pycon talk about native packaging being the main thing to worry about, and single-file binaries being at best a stepping stone to that and at worst a bit of a red herring? You don’t have to take it from me. From the authors of a widely-distributed command-line application that was rewritten from Python into Go specifically for easier distribution, and then rewritten in Python:

... [the] majority of people prefer native packages so distributing precompiled binaries wasn’t a big win for this type of project1 ...

I don’t want to say “I told you so”, but... no. Wait a second. That is exactly what I want to do. That is what I am doing.

I told you so.

Hello lazyweb,

I want to run some “legacy” software (Trac, specifically) on a Swarm cluster. The files that it needs to store are mostly effectively write-once (it’s the attachments database) but may need to be deleted (spammers and purveyors of malware occasionally try to upload things for spamming or C&C) so while mutability is necessary, there’s a very low risk of any write contention.

I can’t use a networked filesystem, or any volume drivers, so no easy-mode solutions. Basically I want to be able to deploy this on any swarm cluster, so no cheating and fiddling with the host.

Is there any software that I can easily run as a daemon that runs in the background, synchronizing the directory of a data volume between N containers where N is the number of hosts in my cluster?

I found this but it strikes me as ... somehow wrong ... to use that as a critical part of my data-center infrastructure. Maybe it would actually be a good start? But in addition to not being really designed for this task, it’s also not open source, which makes me a little nervous. This list, or even this smaller one is huge and bewildering. So I was hoping you could drop me a line if you’ve got an idea what I could use for this.

I like keeping a comprehensive an accurate addressbook that includes all past email addresses for my contacts, including those which are no longer valid. I do this because I want to be able to see conversations stretching back over the years as originating from that person.

Unfortunately this causes problems when sending mail sometimes. On macOS, at least as of El Capitan, neither the Mail application nor the Contacts application have any mechanism for indicating preference-order of email addresses that I’ve been able to find. Compounding this annoyance, when completing a recipient’s address based on their name, it displays all email addresses for a contact without showing their label, which means even if I label one “preferred” or “USE THIS ONE NOW”, or “zzz don’t use this hasn’t worked since 2005”, I can’t tell when I’m sending a message.

But it seems as though it defaults to sending messages to the most recent outgoing address for that contact that it can see in an email. For people I send email to regularly to this is easy enough. For people who I’m aware have changed their email address, but where I don’t actually want to send them a message, I think I figured out a little hack that makes it work: make a new folder called “Preferred Addresses Hack” (or something suitably), compose a new message addressed to the correct address, then drag the message out of drafts into the folder; since it has a recent date and is addressed to the right person, will index it and auto-complete the correct address in the future.

However, since the previous behavior appeared somewhat non-deterministic, I might be tricking myself into believing that this hack worked. If you can confirm it, I’d appreciate it if you would let me know.

I think I’m using GitHub wrong.

I use a hodgepodge of https: and : (i.e. “ssh”) URL schemes for my local clones; sometimes I have a remote called “github” and sometimes I have one called “origin”. Sometimes I clone from a fork I made and sometimes I clone from the upstream.

I think the right way to use GitHub would instead be to always fork first, make my remote always be “origin”, and consistently name the upstream remote “upstream”. The problem with this, though, is that forks rapidly fall out of date, and I often want to automatically synchronize all the upstream branches.

Is there a script or a github option or something to synchronize a fork with upstream automatically, including all its branches and tags? I know there’s no comment field, but you can email me or reply on twitter.