WEBVTT

00:00:00.000 --> 00:00:03.640
I run a podcast. I mean, you're listening to

00:00:03.640 --> 00:00:06.919
it right now. And I genuinely don't know how

00:00:06.919 --> 00:00:11.000
many people heard that last episode. Not approximately,

00:00:11.859 --> 00:00:15.859
not within some margins. I really don't know.

00:00:16.480 --> 00:00:21.460
I get a download count and that's about it. And

00:00:21.460 --> 00:00:25.039
the wild thing is that's not a bug in how I set

00:00:25.039 --> 00:00:27.699
things up. That's just how podcasting works and

00:00:27.699 --> 00:00:30.500
it's... It's been that way since the beginning

00:00:30.500 --> 00:00:34.840
And it traces back to a single XML tag that somebody

00:00:34.840 --> 00:00:37.939
added to a spec on Christmas Day back in the

00:00:37.939 --> 00:00:42.240
year 2000 This is the Tyler Woodward project.

00:00:42.539 --> 00:00:46.000
I'm Tyler all opinions are my own and today I

00:00:46.000 --> 00:00:50.200
want to talk about the RSS enclosure tag What

00:00:50.200 --> 00:00:54.159
it is why it matters and why Spotify spent a

00:00:54.159 --> 00:00:56.659
billion dollars trying to build something that

00:00:56.659 --> 00:01:06.939
would replace it and basically gave up. Okay,

00:01:07.359 --> 00:01:12.019
so RSS, real simple syndication. If you've been

00:01:12.019 --> 00:01:14.480
on the internet since the early 2000s, you might

00:01:14.480 --> 00:01:18.159
remember it. A little orange icon, feed readers,

00:01:18.400 --> 00:01:21.700
Google reader before Google up and killed it

00:01:21.700 --> 00:01:25.840
off. The idea was that instead of going to every

00:01:25.840 --> 00:01:28.750
website manually, to check if there was something

00:01:28.750 --> 00:01:34.430
new you subscribed to a feed your reader checked

00:01:34.430 --> 00:01:38.090
the feed new stuff showed up it was good and

00:01:38.090 --> 00:01:42.930
then it mostly died for regular people once social

00:01:42.930 --> 00:01:46.150
media took over and platforms decided they'd

00:01:46.150 --> 00:01:52.950
rather curate your content for you but RSS didn't

00:01:52.950 --> 00:01:57.519
die in podcasting It became the foundation of

00:01:57.519 --> 00:02:01.900
the whole thing The reason is one little tag

00:02:01.900 --> 00:02:06.799
the enclosure tag Dave winner added it to the

00:02:06.799 --> 00:02:12.400
RSS spec on December 25th 2000 which you know

00:02:12.400 --> 00:02:17.699
is Christmas make of that what you will What

00:02:17.699 --> 00:02:22.199
the tag does is attach a file to a feed item,

00:02:22.199 --> 00:02:26.310
you know, it's normally Normally, an RSS item

00:02:26.310 --> 00:02:30.650
is just text. Title, description, a link. The

00:02:30.650 --> 00:02:34.129
enclosure tag says there's also a file here.

00:02:34.389 --> 00:02:36.310
Here's where it lives. Here's how big it is.

00:02:36.409 --> 00:02:39.389
Here's what kind of file it is. And that's the

00:02:39.389 --> 00:02:43.689
whole invention. Attach a file to a feed item.

00:02:45.090 --> 00:02:48.810
Adam Curry, who was an MTV VJ in the 80s and

00:02:48.810 --> 00:02:52.389
90s, is also a genuine tech nerd. He built the

00:02:52.389 --> 00:02:57.180
script. In 2003, that would look for audio enclosures

00:02:57.180 --> 00:02:59.960
in RSS feeds and automatically pull them into

00:02:59.960 --> 00:03:05.699
iTunes. Somebody coined the word podcasting.

00:03:06.319 --> 00:03:10.020
Apple shipped native support for this in iTunes

00:03:10.020 --> 00:03:16.120
4 .9 back in June of 2005. Get this, a million

00:03:16.120 --> 00:03:20.979
subscriptions in 48 hours. The entire medium

00:03:20.979 --> 00:03:27.150
took off on infrastructure that nobody owns.

00:03:27.569 --> 00:03:32.050
Nobody owns RSS. The specs copyright was assigned

00:03:32.050 --> 00:03:35.129
to Harvard's Berkman Klein Center back in 2003

00:03:35.129 --> 00:03:40.030
under a Creative Commons license. No company

00:03:40.030 --> 00:03:44.289
holds it. You can read the spec right now. Implement

00:03:44.289 --> 00:03:46.530
it yourself. Ship an app tomorrow. No permission,

00:03:46.689 --> 00:03:53.479
no license fees, no no proprietary garbage. That's

00:03:53.479 --> 00:03:58.740
genuinely unusual. The foundational infrastructure

00:03:58.740 --> 00:04:02.159
for an entire industry worth billions of dollars

00:04:02.159 --> 00:04:05.620
in ad revenue and there's no entity that owns

00:04:05.620 --> 00:04:11.319
it or can take it away from you. What that means

00:04:11.319 --> 00:04:15.099
practically is that when you subscribe to a podcast,

00:04:15.719 --> 00:04:20.519
you're just giving an app a URL. The app checks

00:04:20.519 --> 00:04:23.879
that URL for new episodes. When it finds one,

00:04:23.939 --> 00:04:28.540
it downloads the file. Your server delivers it.

00:04:28.800 --> 00:04:32.060
And that's sort of the whole transaction. Apple

00:04:32.060 --> 00:04:34.600
doesn't have to be involved. Spotify doesn't

00:04:34.600 --> 00:04:37.699
have to be involved. Unless you're hosted on

00:04:37.699 --> 00:04:39.920
one of those platforms, the file goes directly

00:04:39.920 --> 00:04:44.959
from your server to the listener's device. So,

00:04:45.060 --> 00:04:47.800
Apple can pull your show from their directory.

00:04:48.000 --> 00:04:50.660
I mean, that happens and it matters because that's

00:04:50.660 --> 00:04:54.019
where a lot of people find the stuff. But they

00:04:54.019 --> 00:04:57.740
can't take your feed down. They can't stop your

00:04:57.740 --> 00:05:00.060
existing subscribers from getting new episodes.

00:05:00.920 --> 00:05:04.120
Compare that to YouTube where your channel gets

00:05:04.120 --> 00:05:08.699
terminated and it's just gone. Or Spotify exclusives

00:05:08.699 --> 00:05:11.959
where the show only exists if Spotify decides

00:05:11.959 --> 00:05:16.519
it exists. The broadcast parallel is actually

00:05:16.810 --> 00:05:19.589
pretty clean here. I mean, think about it, nobody

00:05:19.589 --> 00:05:23.550
owns the FN band. It's public airwaves. The FCC

00:05:23.550 --> 00:05:27.889
regulates it, sure, and can yank your license,

00:05:28.569 --> 00:05:31.829
but the underlying infrastructure is shared.

00:05:33.670 --> 00:05:37.990
Podcasting RSS is sort of similar. The rails

00:05:37.990 --> 00:05:43.129
are shared. You just have to host the files somewhere.

00:05:44.490 --> 00:05:49.540
Okay, but Here's the thing. I can't get over

00:05:49.540 --> 00:05:54.100
I Said I don't know how many people actually

00:05:54.100 --> 00:06:00.139
listen to my last episode that's because in podcasting

00:06:00.139 --> 00:06:04.139
a Download is an uh, you know an HTTP request

00:06:04.139 --> 00:06:09.399
your server sees an IP address a user agent a

00:06:09.399 --> 00:06:13.839
timestamp and how many bytes get served from

00:06:13.839 --> 00:06:19.100
the server It doesn't see whether anyone hit

00:06:19.100 --> 00:06:21.600
play. It doesn't know if the file set on a phone

00:06:21.600 --> 00:06:23.860
for three weeks and then get deleted because

00:06:23.860 --> 00:06:27.980
it was never played. There's no session. There's

00:06:27.980 --> 00:06:32.939
no callback when someone finishes. In radio,

00:06:33.120 --> 00:06:35.939
and I've worked in radio for a little while now,

00:06:36.439 --> 00:06:39.980
we have Nielsen, PPM meters, diary surveys, panels.

00:06:41.860 --> 00:06:45.439
It's imperfect, but... PPM has well -documented

00:06:45.439 --> 00:06:50.160
sample size problems in small markets, and if

00:06:50.160 --> 00:06:52.699
you've ever looked at the margins of error in

00:06:52.699 --> 00:06:56.680
a Nielsen book, you've had a small existential

00:06:56.680 --> 00:07:01.680
crisis, but you know at least something. You

00:07:01.680 --> 00:07:07.160
have a number you can defend to an advertiser.

00:07:07.720 --> 00:07:11.759
Podcasting has the IAB TechLab podcast measurement

00:07:11.759 --> 00:07:15.730
guidelines. Try saying that five times fast.

00:07:16.589 --> 00:07:21.649
Four versions now. The latest is from early 2024.

00:07:23.149 --> 00:07:26.449
What they define is a process for cleaning your

00:07:26.449 --> 00:07:30.649
server logs to get a more defensible download

00:07:30.649 --> 00:07:35.069
count. De -duplicate by IP and user agent within

00:07:35.069 --> 00:07:39.850
24 hours and force a minimum byte threshold filter

00:07:39.850 --> 00:07:45.529
out known bots. What you get at the end of that

00:07:45.529 --> 00:07:48.730
process is a number that more accurately represents

00:07:48.730 --> 00:07:54.790
deliveries not listens deliveries again downloads

00:07:54.790 --> 00:07:59.290
Spotify knows your age bracket. They know exactly

00:07:59.290 --> 00:08:01.529
where you dropped off on an episode They know

00:08:01.529 --> 00:08:05.009
if you skip the ad that's because they own the

00:08:05.009 --> 00:08:11.000
client in the server They see everything If a

00:08:11.000 --> 00:08:14.720
platform owns the whole stack, you get real analytics.

00:08:15.519 --> 00:08:18.379
If you want independence from the platform, you're

00:08:18.379 --> 00:08:22.279
working from server logs. Spotify understood

00:08:22.279 --> 00:08:26.500
this, which is why they started back in 2019.

00:08:26.779 --> 00:08:30.779
They went on a massive spending spree. Gimlet

00:08:30.779 --> 00:08:35.200
for around $230 million. Anchor for around $154

00:08:35.200 --> 00:08:38.600
million. Parcast, The Ringer, Megaphone, Podsights,

00:08:38.779 --> 00:08:41.740
Chartable. Joe Rogan went exclusive in December

00:08:41.740 --> 00:08:47.960
of 2020 for a reported $200 million plus. Total

00:08:47.960 --> 00:08:53.460
spent on the entire beta test cleared billions

00:08:53.460 --> 00:08:57.759
of dollars. The bet was that if they could pull

00:08:57.759 --> 00:09:01.879
enough listeners into Spotify with exclusive

00:09:01.879 --> 00:09:04.659
content, they could migrate the medium off of

00:09:04.659 --> 00:09:09.600
open RSS standards and onto something they controlled.

00:09:11.059 --> 00:09:14.519
It didn't work. The Obama production deal ended

00:09:14.519 --> 00:09:18.100
and went to Audible for wide open distribution.

00:09:18.580 --> 00:09:21.860
Brene Brown's show left exclusivity and moved

00:09:21.860 --> 00:09:28.039
over to Vox Media. Rogan's renewal in early 2024

00:09:28.039 --> 00:09:34.659
was explicitly not exclusive. He went back to

00:09:34.659 --> 00:09:37.419
Apple, YouTube, Amazon Music all on the same

00:09:37.419 --> 00:09:42.549
day. Spotify said it was the last exclusive podcast

00:09:42.549 --> 00:09:47.210
deal they'd ever do. One thing worth knowing,

00:09:47.549 --> 00:09:52.929
when Spotify bought Podsites and Charitable in

00:09:52.929 --> 00:09:56.429
2022, those were the only two third -party measurement

00:09:56.429 --> 00:09:59.950
platforms that worked across publishers and advertisers.

00:10:01.350 --> 00:10:04.330
Charitable got folded into megaphone and shut

00:10:04.330 --> 00:10:11.149
down. So Spotify The dominant non -apple listening

00:10:11.149 --> 00:10:15.690
platform now also owns a big piece of the neutral

00:10:15.690 --> 00:10:17.970
measurement infrastructure everyone else was

00:10:17.970 --> 00:10:23.649
using. RSS didn't fix that. That's a platform

00:10:23.649 --> 00:10:27.549
layer problem. The protocol has no mechanism

00:10:27.549 --> 00:10:31.070
to address. The other thing I find interesting

00:10:31.070 --> 00:10:35.809
is podcasting 2 .0. Adam Curry and a developer

00:10:35.809 --> 00:10:39.309
named Dave Jones started it in 2020. It's an

00:10:39.309 --> 00:10:44.629
open extension to the RSS namespace, so it layers

00:10:44.629 --> 00:10:47.549
on top of the existing protocol and doesn't replace

00:10:47.549 --> 00:10:51.909
it. It adds transcripts, JSON chapters, direct

00:10:51.909 --> 00:10:54.490
micro payments via the Lightning Network, live

00:10:54.490 --> 00:10:59.570
streaming support, globally unique episode IDs.

00:11:00.549 --> 00:11:05.730
Apple shipped support for it. and iOS 17 .4,

00:11:05.850 --> 00:11:09.610
at least for the transcript part. The first time

00:11:09.610 --> 00:11:12.590
they adopted an open namespace tag rather than

00:11:12.590 --> 00:11:16.110
inventing their own proprietary version. Spotify

00:11:16.110 --> 00:11:18.649
drugged their feet for a little while, but added

00:11:18.649 --> 00:11:23.909
it later that year. The mainstream apps are moving

00:11:23.909 --> 00:11:26.970
slowly. Nobody's done the micropayment stuff

00:11:26.970 --> 00:11:31.090
at scale. but the direction is keep the open

00:11:31.090 --> 00:11:34.350
protocol, add structured features on top of it,

00:11:34.870 --> 00:11:39.590
and don't try to instrument the listener. That's

00:11:39.590 --> 00:11:46.129
a choice. A download is not a listen. It's never

00:11:46.129 --> 00:11:49.730
been a listen. The entire podcast medium has

00:11:49.730 --> 00:11:54.870
been running on that vagueness since 2005, and

00:11:54.870 --> 00:12:01.100
20 years in, It's still not resolved. I've been

00:12:01.100 --> 00:12:03.879
thinking about this because I'm literally on

00:12:03.879 --> 00:12:07.039
both sides of it. I work in broadcast where the

00:12:07.039 --> 00:12:11.919
measurement is imperfect, but at least something

00:12:11.919 --> 00:12:16.299
exists. And I run a podcast where I get a download

00:12:16.299 --> 00:12:18.980
count and I'm supposed to generalize from there.

00:12:20.240 --> 00:12:24.899
The same design choice that makes this show impossible

00:12:24.899 --> 00:12:31.259
to the platform is the one that makes it impossible

00:12:31.259 --> 00:12:35.179
to fully measure either. Not really separate

00:12:35.179 --> 00:12:38.399
problems, same decision and I'm not sure I'd

00:12:38.399 --> 00:12:42.320
trade it because I've watched what happens when

00:12:42.320 --> 00:12:45.940
a platform owns the whole stack and decides the

00:12:45.940 --> 00:12:53.370
economics don't work anymore. The billion -dollar

00:12:53.370 --> 00:12:57.690
siege failed. The 24 -year -old XML tag is still

00:12:57.690 --> 00:13:02.669
running, still doing its thing. Thanks for listening

00:13:02.669 --> 00:13:06.649
to the Tyler Woodward Project. This is an independently

00:13:06.649 --> 00:13:09.990
produced and written podcast by myself, and if

00:13:09.990 --> 00:13:11.950
you'd like to support the show and project, head

00:13:11.950 --> 00:13:15.450
over to tylerwoodward .me forward slash subscribe

00:13:15.450 --> 00:13:18.970
and become a member today. It goes a long way.

00:13:19.100 --> 00:13:21.960
Makes me feel all loved and stuff a little warm

00:13:21.960 --> 00:13:25.460
and fuzzy inside. You can also follow me on Instagram

00:13:25.460 --> 00:13:30.399
threads and blue sky at Tyler Woodward dot me

00:13:30.399 --> 00:13:32.340
I'll catch you guys next week