Some have noticed a brief outage earlier today. That was me, moving OnePo and friends to a new server, running a new long-term-support OS.

We're on Debian 12 Linux now, for those who care about such things. Supported until mid-2028! A genuine, democratically run community project!

In mid-2028 I may even be able to upgrade in place, although it's usually a good practice to redo these things from time to time.

This all went much more quickly than it might have. Like... two hours of knocking around? Tops? I had minimal trouble moving my nginx, certbot, postfix and opendkim configurations over as-is.

[Knocks on virtual wood]

I also got a slight price cut out of the deal.

Still running on Linode. I considered moving to AWS Lightsail, but Lightsail is just a less convenient clone of Linode for people whose bosses insist on AWS, and I'm not one of those bosses. Except when I'm at work. Linode offered shiny-modern Debian 12 as a standard option, and the ability to just swap my existing IP address to the new server as soon as it was ready to go. Bamfl!

I'm keeping the old server around, powered off, for one week juuuuust in case I missed something.

A user has asked if I would consider adding direct messaging to OPW. Hmm.

I can think of one constraint on that feature right away: the bar to establish your right to DM someone needs to be pretty high. Like maybe you have to be mutual followers to start with, and then you can request DM access from that person. Even then I'm not sure how I feel about it. It's certainly not posting once a day. But it has a lot of important applications in situations that call for compassion, rather than being a stickler for the "spirit of things." Hmm.

Private comments to specific responders on a particular post have also been mentioned as an idea. Perhaps you, as the original poster, would be able to reply to a commenter privately.

That  stays closer to the idiom of One Post Wonder, but I think that it could generate confusion, e.g. accidental references to things said in private comments in public comments on the same post.

What do you think?

I fixed a lil' bug that caused comments to contain HTML markup tags if edited again later. Thanks to both users who reported this, and apologies for not figuring out how to reproduce it the first time.

Unfortunately, in writing this post I encountered a second bug I can't fix before work this morning. It affects the second rich text block in the same post. Sigh. Will dig into that. These are ripple effects of recent modernization of the codebase, which I can't regret really. There is no other bug. I ignored my own "Alternative Text" placeholder and confused the alternative text field with a new rich text block because I am not awake yet. As an amateur UX designer I can see how that confusion happens, so I might need to more clearly flag these.

Two pictures from the American Museum of Natural History: one from the new Invisible Worlds exhibit, one compelling traditional exhibit. If you check out Invisible Worlds, be sure to pay attention to how the little kids interact with the pressure-sensitive floor. You can pay attention to the Serious Content For Grownups on the second loop.

100 feet down in a former copper mine. Which is also a former prison. The past is a terrible place, never go there. Old Newgate Prison is worth a visit, though. The curation of the place... nowadays... has tipped heavily in the direction of rethinking incarceration. Also: nice vein of malachite in there.

Speaking of the past, thanks to U. and L. for bringing some nasty bugs to my attention. These bugs affected the process of accepting an invitation. Most were blatant, which is not so bad, because it was easy for L. to spot them and for me to fix them.

But the last one makes me kick myself: for *coughs* some time now, when accepting an invite, the dialog box giving you a chance to give a "friends" key to the person who invited you just didn't work properly. This is fixed, and will hopefully make it just that little bit easier for newcomers to communicate with the person who invited them in the first place.

Thankfully there was no issue in the opposite direction.

Is OnePo fully accessible? Um... yeah almost certainly not, because I haven't had it tested rigorously for that, nor have I consulted anyone with a disability about their actual experiences. Those steps would be good to take.

One bit of accessibility progress though: alternative text for images.

Adding an alt text field for images in posts was easy-peasy. Adding this to images in comments was an absolute bear, because of the lil' shortcut pattern I had before: tapping the image button on a comment immediately prompted you for an image file, and then immediately uploaded it, and you were done!

This was a nice pattern in a lot of ways, very light and breezy, but also totally brittle: the moment you wanted to add alt text it didn't work for us at all.

So I reworked the image comment editor, and now there is an alt text field beneath an "Upload Image" button, all proper-like.

There's another win today though: image uploads should be considerably faster, especially over slow links, and especially when the original file is large. That's because images are now resized on your device before they are uploaded.

It's true that this puts a slightly greater burden on the local device, but in practice it's a device every web browser has been ferociously optimized to handle, and most likely handles in dedicated hardware. That's because every website is crammed to the gills with images that don't precisely match the size of your screen. So asking your device to do this is reasonable, and it significantly reduces the workload for the server.

"Uh... yeah Tom, that workload must be awful with somebody posting once every hour or so!" Yeah, true, but I still get to learn useful things about doing this kind of stuff at scale, even if scale hasn't really happened here.

Anyway, here's Wonderwall. Photo-comment away, and let me know if anything breaks!

P.S. Hovering the mouse etc. will not display the "alt text." That's because it is truly set up as alternative text (e.g. the "alt" attribute) for accessibility purposes, and regular browsers intentionally don't display it by default. That "text on hover" thing you may be thinking of is the "title" attribute, and describing an image effectively for those who can't see it isn't really the same use case.

I made some improvements to OnePo this week. Specifically:

• I brought back the ability to insert more text, image and video blocks in between existing blocks in your post. Personally I found it too crazymaking to always add things at the bottom and then nudge them up.
• However, there were user experience concerns expressed about that being too much before. This is why I took a stab at a better experience for it: the buttons to add new blocks are nestled in a single "+" toggle above each existing block, and you can toggle it closed again.
• Speaking of nudging things up from the bottom, that was not a great experience either, particularly if you had to nudge something with the arrows ten times. So now we drag blocks around instead.

These changes work well on desktop, should also work well on tablets (let me know if you get the chance to try it), and work okay on mobile. I do find the new experience of dragging an image a little weird on the iPhone, but it works and hopefully I can continue to fine-tune it.

However... these changes are a bit selfish. There's nothing wrong, obviously, with putting my own uncompensated free time into features I like. Honestly! The nerve of this guy!

But, I've also promised alt text for images and that's a matter of accessibility. So I will get back to that next.

Alt text will be easiest to add for images in posts, because those are full-blown blocks in the editor. Adding one more editable field to them is no big deal.

But for images in comments, it's necessary to rethink the user experience. Right now it's a bit of a hack: "just pick an image and it'll upload and bam your comment is live!" That's cute, but it closes the door on adding alt text before your comment is live. And also there is currently no way to edit an existing image comment. So yeah, rethink.

I did carry out some interesting experiments in that area recently, which also included improvements to the image uploading experience. If I manage to carry that through, image uploads will be a lot faster because the browser will be doing the initial resizing work if necessary.

I fixed a bug. Turns out "delete comment" didn't work, for some unspecified amount of time, but probably since early this year. It failed in a very obvious way though, so I doubt anyone was misled into thinking they had removed a comment. All tidied up, sorry about that. Thanks to the person who reported it.

I also modernized a few boring technical details while I was in the neighborhood.

Every now and then I remember my new machine has a touchscreen. Then I try to forget. I'll never clean it.

No more goldang inexplicable sidey-scrolling on yer phonez goldangit!!

What was it, you ask?

Why, it was the fleuron, of course. The fleuron with the negative margins.

He wants a fleuron

With negative margins [doop dweet doooo]

Because he knows there's no such thing

Fixed:

• The "cancel" button for new, never-saved comments works reasonably now (broken for a long time).
• Pasting a link to something well-annotated for social sharing, like a CNN article, now generates a nice "open graph" summary again. Also probably broken for some time super broken since ever in half a dozen ways, sorry. Debugged the daylights out of this.
• Image comments (broken by the recent rewrite).
• Layout issues on the iPhone, especially when open graph is in play.

Also, code quality things:

• Removed yet more "dead code" (code no longer in use, exactly like junk DNA: if left in place too long there's a chance it'll be "in use" after all, in some sneaky and incomprehensible way).
• Stopped using the obsolete "request" module to fetch stuff.

All in all, I'd say I got away with the rewrite pretty well. Nothing went hugely sideways. Twitching to do more with the site now, which is a nice feeling after slogging through / dreading the rewrite for so long.

Thanks to all who reported bugs.

Nice Open Graph example (CNN article).

Vastly more important Open Graph example (Mastodon #caturday post).

Looks like Mastodon's og:image tags have site-relative URLs. I noticed 'cuz I had a bug handling those. Amusingly, Facebook does not support relative URLs in open graph at all. This means I now have better support for Open Graph than Facebook. Phbbbbbt.

BTW Open Graph only "unfurls" if you paste a link straight into your text without clicking the link icon in the rich text editor. So if you don't want it, if you just want to type some text and then link it to something because this is the hyper-texter-webs goshdarnit, then type that text, select it and click the link button.

Open Graph still needs some styling love on mobile fixed that shizzle too.

Hey, did you feel something? I hope not.

I've deployed a new version of the One Post Wonder source code. And it should be... exactly the same, for you. If not, please let me know.

For avoidance of confusion: if something seems busted, it's not you, it's me. Let me know. Thanks! I did lots of testing. But... software is hard.

Just in case, my email is tommybgoode@gmail.com. If stuff broke, tell me things ASAP. Thanks!

What Changed

"So... if everything is the same... what's different?"

I rewrote the code using a modern idiom that other programmers will be able to follow. Which makes it more practical for others to help me move OPW forward. It's a hell of a lot easier for me to maintain the code as well.

You see, back in the Bad Old Days, Node.js programming was like this:

"Hey Jane, fetch me some blog posts, then call me." [Hand-sign of phone]

"Hey Jack, when Jane calls me, fetch some comments, then call me." [Hand-sign of phone]

"Hey Joe, when Jack calls me after Jane calls me, fetch some images, then call me." [Hand-sign of phone]

Except dozens of requests are coming in at once so it's like a telethon in there with only one person manning the phones and everybody wants a fucking tote bag and theoretically that's OK because you don't actually speak continuously to one person right, so...

AAAAAAAA yeah complete madness.

In the New Idiom, Node.js JavaScript programming is like this:

MOTHER MAY I fetch some blog posts.

MOTHER MAY I fetch some comments.

MOTHER MAY I fetch some images.

Behind the scenes, it's the same telethon, but the management of all that parallel madness is encapsulated in the magic words "MOTHER MAY I," which allow us to express what we need to do as a simple series of steps and let JavaScript sort it out behind the scenes.

An astute observer will note that I still have to say "MOTHER MAY I" before each thing. I can't just do the fucking things. Which is still weird and unintuitive. However, in modern JavaScript "MOTHER MAY I" is expressed in the single word "await," which is at least convenient and readable. Crucially, it's at least readable by people who don't do this all day and who don't know this particular codebase by heart.

And the performance benefits are real enough that these days, Ruby, Python, PHP and even Rust optionally offer the same thing for those who need the speed.

So, we've got that going for us.

[Waves tattered flag]

Wait. Why do we do this?

A person might ask why programmers put up with any of this. Java and Python and Ruby and frickin' PHP all let you just write:

Do a thing

Do another thing

Do a third thing

THE END

So why would you put up with this "MOTHER MAY I" bullshit from JavaScript?

Well... because it's fast. Most languages for the web give you a separate thread, aka basically a completely separate telethon volunteer, to think about each request, aka person calling to get a tote bag. Which is enormously convenient. But, also slower because you're not maximizing the throughput of your telethon volunteers.

And... you have to write JavaScript for the browser anyway. To a first approximation, at least. So you may as well double down and use those skills on the server too, right? Less mental context switching. Which gives you plenty of energy for typing MOTHER MAY I! This is sane. [Nods firmly]

Um... great! What's next?

"So what's next?" New features? Probably new features. I promised alt attributes a LONG time ago.

Probably also breaking up that massive "routes.js" file into separate collections of API routes by topic, so I can find stuff.

Probably also TypeScript, just so I can put "8 billion years TypeScript experience" on my resume.

Thank you sir, may I please have another?

"What's TypeScript?" It's a new flavor of JavaScript that is "strongly typed." That means you have to say "MOTHER MAY I" even more often and in twisty new ways... but, if you fail to say it in exactly the same way when you try to call that same code later, it'll stop you and say "uh-uh, that's going to introduce a bug."

So it's a good thing.

One may well ask: does that mean TypeScript can check the original "MOTHER MAY I" incantations (aka "await" keywords) to make sure none are missing? That would be seriously useful.

Oh gosh no, it can't do THAT.  Who would want to catch by far the most common source of bugs in a Node.js program? No, it only stops you if you pass a string instead of a number (*), which I practically never do.

But I hear you can do it if you combine TypeScript and eslint and a roll of duct tape. So maybe I'll give that a try.

(*) Yes, yes, and a thousand more complicated but ultimately similar things.