inessential by Brent Simmons

The Developers Union

Some of the press coverage about The Developers Union uses words like “angry” and “fed up.” These aren’t accurate characterizations at all. Nobody’s mad here!

But here‘s the deal: Apple controls the App Store and its economics. The system could be set up better to support high-quality apps, by indies, that last for years.

Apple doesn’t have to, of course. But we can ask! It’s totally okay to ask, so we are.

We think that an important first step would be a standardized, App-Store-supported way of offering free trials. (And where, once purchased, Family Sharing works.)

Trial versions have worked great for years for indie Mac developers, before the App Store, and we think it would benefit indies on the iOS and Mac App Stores.

And the platform would get better — and more sustainable — apps. Everyone wins!

If you agree, you can sign up. Add your name. Add your app.

I realize you might be worried about doing a thing that could upset powerful people inside Apple. I strongly doubt that that worry is actually well-founded — but, then again, that’s part of why this is a big list.

* * *

I should note that I’m not doing this as part of Omni. I’m not even doing it for my side projects — they’re all free, and it’s quite possible that none of them will ever appear on any App Store at all.

Instead, I’m thinking of my friends, of developers I admire, of up-and-coming developers I haven’t even heard of yet. I — quite selfishly! — want them to thrive. I want to see what great stuff they could make. I want everybody to have the opportunity I’ve had.

I’ve been lucky, and I’ve done well — and my experience should not be rare.

OmniFocus 3.0 for iOS ships in four weeks.

As Marketing Human, I’ve got work to do! But I’m totally psyched.

Making Apps Is Harder Than It Needs To Be

With the recent talk about Electron and “Marzipan” — or maybe Amber or something, according to Mark Gurman — I’m reminded of a thing I think about kind of often: that making iOS and macOS apps is way harder than it needs to be.

For most apps (except games, I suppose), a huge percentage of the code might as well be written in a scripting language. We absolutely do not need to be writing everything in Swift, Objective-C, C++, or C.

“But Brent,” you say, “what about performance?”

Consider the case where you set up an animation and then run the animation. The system does that animation. Or consider Core Data — your choice of language doesn’t affect how fast it can read from SQLite. Or think of networking — it’s bound by the connection, not the speed of your code. Or think of pushing a view controller onto the current navigation view controller. Or setting up view constraints. And so on.

All this code might as well be Ruby — or, preferably, a scripting language designed for app making. (I would have liked an Objective-C-without-the-C.)

And the thing that would make it all so worthwhile is editing the code while the app is running. You could go all day without an explicit build step!

Sure, some of your code would still have to be written in Swift or whatever. The part that really does have to be fast. I’m a performance junkie myself, so I get this. (Evergreen’s RSS parser is fast, and I wouldn’t switch it to a scripting language.)

But most of most apps (again, probably besides games, about which I know nothing) could be written using a scripting language.

PS Yes, I’m quite aware that we used to have Fix & Continue. And WebScript.


I wouldn’t wait for “Marzipan” or XKit or whatever it is.

We don’t know what it is. But my guess — based on my 38 years of writing code for Apple computers — is that it’s something you can use along with UIKit and AppKit, and not a wholesale replacement.

Maybe it’s a declarative API that helps make some things easier, and maybe you can make a cross-platform button more easily. Maybe your table view code could be the same on iOS and macOS. Great!

But don’t expect Macs to turn into large iPads all of a sudden. Macs are gonna Mac. Apps are going to have multiple resizable windows and a menubar. Targets will still be sized and designed for mice and trackpads.

In other words, if you want to write a Mac app, you’re still going to have to deal with the things that are inherently different about Mac apps, regardless of the specific API.

Let’s say this thing ships in the fall of 2019, over a year from now. If past is a guide, we might imagine it would be fun to play with, but not more useful than, say, the original version of Swift. (Swift didn’t get really good for writing apps until Swift 3.)

So it might be 2020 before it’s something that accelerates Mac development in any real way.

You could write a few Mac apps between now and then.

* * *

I realize that documentation on writing Mac apps is hard to find these days. Books on the subject are rare, and any book you find may be out of date.


One of the reasons I made Evergreen open source is so that people who want to write a Mac app have some examples.

And I just learned that there’s a big list of open source Mac apps. This is way more than than was available when I started writing apps for OS X.

I don’t have time to write a book on Mac app development. I wish I did. I might make the time to do a small article now and then, using Evergreen as example. Maybe.

But it’s not my job (as I have to keep reminding myself). It’s Apple’s job to document and evangelize the Mac platform.

(As an additional part of that, I’d like to see Apple update the Mac App Store, and maybe also deal with some of the issues with sandboxing. It would signal that the company cares about Mac apps. I know it does care, but a more public demonstration would be welcome.)

Check out Automation Orchard, a new site by Rosemary Orchard that is the “place to find resources to help you automate your life.”

I immediately thought of ScriptWeb — which, to my delight, is still up! Though its last update was 2009.

Evergreen/Frontier Status: ODB Work

For the past few days I’ve been working on adding Frontier-like object database (ODB) support to my database framework.

It’s not finished yet — it doesn’t even build.

What it is (or, what it will be)

It’s hierarchical key-value storage. No schemas. Tables can contain tables, with no limit.

This implementation is the lowest level: the part that gets, sets, and deletes data from the database.

It’s application-agnostic, at this level — it doesn’t know about all of Frontier’s data types, for instance. A level on top of this will be needed for new-Frontier.

SQLite, my favorite hammer

I’m not actually writing a new database — I’m using SQLite. And that’s because I’ve been using SQLite for 15 years, and I love it and know it well, and I know how incredibly stable it is. I’m not willing to write my own thing, and I’m not willing to use a thing less mature and rock-solid than SQLite.

How it works:

The schema is pretty simple. There are tables and values.

Every table has an id. Every table (except the root table) has a parent_id that points to its parent table.

And every value has an odb_table_id that points to its parent table.

This way it’s easy to get a table’s children: it takes just two select statements.

(Both tables and values also have a name, since this is key-value storage.)

Tables and values will be cached in memory, so not every call will require a database read.

(Before you suggest I use something other than SQLite, know that I won’t change my mind on this.)

(Also, again: it’s not done yet. Doesn’t even build.)

Why I’m doing this now instead of something else

I’m using schema-less storage for feeds in Evergreen. (Articles and article status, on the other hand, are stored using a schema, in SQLite.)

Currently I’m writing a big binary plist with all the feed data, and it has to be rewritten every time a feed property changes. The writes are coalesced — but still, this isn’t great.

I’m using schema-less storage in part because of syncing systems: I don’t know, and can’t guess, what I’ll need to store. Different systems will have different requirements.

Also: I may add features later that require additional feed properties. I don’t know what those are.

I realized that what I really want for this is a feature from Frontier: hierarchical key-value storage.

Each system will gets its own database on the client. For each, I’ll create an odb table called feeds. Each feed will have its own subtable. The key will be its id (which may or may not be its URL, depending on the syncing system).

And inside each subtable I can put whatever I want, at any time, without having to change any schemas or implementations.


For the On My Mac account — not synced; reads feeds directly — we keep track of Etag headers in order to support conditional GET. So, for example, I’d want to get, set, and delete feeds.[feedID].etag.ifModifiedSince.

But with most syncing systems we get the feed content from the system itself — not by directly reading the feed. There might be some other data from the service to store: feeds.[feedID].syncToken, for instance.

You’re Practically a Mac Developer

Say you write an iOS app, and now you want to write the Mac version.

Assuming there’s a data model, maybe a database, some networking code, that kind of thing, then you can use that exact same code in your Mac app, quite likely without any changes whatsoever.

That leaves the 20% or whatever that’s user interface. AppKit is not the same as UIKit, but it’s recognizable. Same patterns and concepts, and often similar names (UITableView/NSTableView).

Given that you’ve done the hard thing — learning UIKit, Xcode, and Swift and/or Objective-C — taking the next step and learning AppKit seems like a very small thing. You’ve climbed the mountain already, after all.

You might complain that AppKit has some weird stuff. True. Some of it, though, isn’t truly weird — it’s just weird to you if you’ve never dealt with things like a menubar and multiple, live-resizable windows.

People coming from AppKit to UIKit (few people these days; many people 10 years ago) might also complain about safe content area insets (or whatever the thing is these days) and size classes and all manner of strange stuff they like not having to deal with in Mac apps. UIKit’s weird too, to some people.

Ten years ago I thought that all the new iOS developers would translate to lots more Mac developers. That that didn’t happen is a huge surprise to me. Because if you’re an iOS developer you’re practically a Mac developer already.

(And — little-known secret — the economics of Mac apps appear to be more favorable than for iOS apps.)

The latest episode of The Omni Show is a special episode — we talk about OmniFocus 3 and flexible inspectors, enhanced repeating tasks, batch editing, and the interleaved Forecast view.

Regular interview shows are our bread and butter, but these roundtables are fun to do too. (And I can’t wait for The Omni Show Live next door to WWDC!)

There’s an unofficial Seattle Xcoders this Thursday at the Cyclops in Belltown. I plan to get there around 6 pm.

We’re always in back, next to the bar but technically in the restaurant section. Anyone is welcome — you don’t have to be a coder! We regularly have designers, testers, support people, product managers, and so on.

Heck, even if you’re a fan, you should come. Should be a beautiful night to hang out with some fine folks.

I was happy to read that Unread 1.9.3 now handles untitled posts better. Very cool.

On the Omni blog I wrote up how we do The Omni Show.

The post explains my approach to marketing, unchanged over the decades:

I don’t have some grand marketing philosophy, other than 1) make great apps, and 2) look out and let other people look in.

Now I’m in a Pickle with this Web Stuff

To publish to this blog, I run a little web server on my Mac that implements the MetaWeblog API, which then renders this blog and rsyncs it to the server. (This way I can write using MarsEdit.)

What I’d rather do: run that little web server on the actual server, and do the static-site generation there. That way I can post from my iPhone and iPad, not just from my Mac.

But… here’s where web deployment gets tricky. I’m on an inexpensive shared host plan at DreamHost. The machine is running an older version of Ruby that’s incompatible with my scripts.

I could use RVM and Bundler, I guess, to use the version of Ruby I want to and to install the gems I need. (It’s just a few, but it’s more than zero.)

That is, if I could figure out how to use this stuff and get it installed on the server. Looks like something I could spend weeks doing (remember that my hobby coding is limited to nights and weekends).

Alternately, I could get an inexpensive VPS from one of the various providers and set things up there. That might be easier — maybe I could skip RVM and Bundler and just install the things I want to use in the old-fashioned way.

But then I have to deal with a bunch of other things myself, including setting up Apache or Nginx. All the things DreamHost does for me automatically I’ll have to handle myself. That doesn’t sound like fun at all.

I totally don’t know what to do. It’s not my plan to become a Ruby deployment expert or to be on the hook for running a server all the time. I’ve done way too much of that kind of thing for one lifetime already, and I’ve mostly been glad to be out of it.

What surprises me is that in 2018 it still requires so much work just to get a CGI script running on a server. It should be easier.

Laura Savino explains the difference between optimal compiling and compiling with optimizations — and which Swift flags mean what.

On the blues harp:

A diatonic harmonica is designed to ease playing in one diatonic scale…

Blues harp subverts the intention of this design with what is “perhaps the most striking example in all music of a thoroughly idiomatic technique that flatly contradicts everything that the instrument was designed for.”

Jason Kottke reminds us that blogging is most certainly not dead, and that there are great blogs out there.

My only objection is the use of the word “dead” to apply to things that aren’t alive. Even when you’re saying that something is not dead.

I’ve done it myself. It’s shorthand, yes, but it’s a broad binary take when something more nuanced and true would be warranted.

The View-Source Web

A line in Frank Chimero’s article Everything Easy Is Hard Again, published a couple months ago, has stuck with me:

That breaks my heart, because so much of my start on the web came from being able to see and easily make sense of any site I’d visit. I had view source, but each year that goes by, it becomes less and less helpful as a way to investigate other people’s work.

One of the ironies of this is that HTML5 makes it easier than ever to make readable, simple HTML. I especially like two things:

  1. Quotes for attribute values are optional (when there are no spaces), and
  2. There are semantic tags for things where before you had to guess at the author’s intention. We have header, main, nav, article, and similar now.

I realized that this blog — since it doesn’t use cookies or JavaScript, since the layout is as straightforward as can be — would make a good personal test case. How easy-to-read can I make the HTML?

So I adopted the semantic HTML5 tags, simplified a few things, and now the source is as easy to read as any HTML I’ve ever written.

Lesson learned: the discoverable and understandable web is still do-able — it’s there waiting to be discovered. It just needs some commitment from the people who make websites.