lee/hpr-knowledge-base
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial commit: HPR Knowledge Base MCP Server

Browse Source 
- MCP server with stdio transport for local use
- Search episodes, transcripts, hosts, and series
- 4,511 episodes with metadata and transcripts
- Data loader with in-memory JSON storage

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Lee Hanken
2025-10-26 10:54:13 +00:00
commit 7c8efd2228
4494 changed files with 1705541 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

440
hpr_transcripts/hpr3648.txt Normal file
View File

@@ -0,0 +1,440 @@
Episode: 3648
Title: HPR3648: A response to tomorrows show
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr3648/hpr3648.mp3
Transcribed: 2025-10-25 02:48:23
---
This is Hacker Public Radio Episode 3648 for Wednesday the 27th of July 2022.
Today's show is entitled A Response to Tomorrow's Show.
It is hosted by Ken Fallon and is about 28 minutes long.
It carries an explicit flag.
The summary is, Ken brings the DeLorean up to Earth to address Monochromex comment
on stats.
Back to the future.
Hi everybody, my name is Ken Fallon and you're listening to another episode of Hacker Public
Radio.
In Tomorrow's show, the Linux in-laws travel back in time to bring us reports from the
future.
Unfortunately, they took a left turn down the wrong leg of the trousers of time and ended
up making some wrong assumptions about how popular they are.
Although the entire show is a spoof based around their meteororic rise to him unfortunate,
it's a segment 6 minutes 31 to 10 minutes 56 that I want to discuss.
It more or less comes down to the following quote.
But if I take a look at ARCA or if we take a look at ARCA for the last one year or almost
almost one year and a half, we clock in on average between 1500 and 2500 listeners.
Given the fact that we have launched this podcast, short of two and a half years ago,
that's quite amazing.
And then later, on average, we are listened to by anything between 5000 and 10,000 business
per episode.
Given the fact that, as I said, quite a few people syndicate us, we're just pointing
the right place.
So you guys, are you sure you got the decimal point in the right place?
Maybe I'm off by magnitude, so maybe just 50,000 to 100,000 people.
So the logic employed here is tick the downloads from one site, multiply that number by this
number of syndicated sites, and that will give you the total downloads.
Now I think we can do a lot better than a mere 100,000.
So first thing, let's look in the Hacker Public Radio logs.
For example, episode HPR 3609 is their latest one.
Linux employs 0.01, episode 57, operating system level virtualization and martens fit.
So a simple grip, dash i, six, three, six, zero, nine, asterix.log, and pipe that to WC,
dash l, or account, that'll give us a total of 8,421, and we do a quick Google search
for Linux outlaws, and that turns 564,000 results in 0.54 seconds, no less.
So if you multiply the number of hits on HPR by the number of results you get in Google,
you arrive at an estimated listenership of 4,749,444,000, now let's round that up to
around 5 billion shall we, and say that's pretty impressive, but we can go further, because
that's just from one show.
We've already released 57 episodes, so the listenership has to be 57 times greater.
So obligatory drumroll, and you can logically say that the Linux in law show has a total
of 270 billion, 718 million, 3,000, and 8 subscribers.
Let's round that up to an even 271 billion shall we.
Now that is impressive, given the fact that the total number of people who lived on planet
earth ever is 108 billion.
So all messing aside, there's something wrong with my logic definitely, but the question
is, is there something wrong with their logic?
Separating the wheat from the chaff.
When I searched through hacker public radio logs earlier, it returned 8,421 hits, but the
internet archive only shows 1,493 downloads.
So what's going on?
Well you guessed it, our logs contain a lot more than just download records.
We need to limit ourselves to counting the media for a start, and that reduces us by 3,713
log lines.
For those interested, gone are 2,169 references where the number 3609 appeared.
For example, 5, 6, 3, 6, 0, 9, 6 bytes in a log line number.
There were 1,107 hits to the episode page itself.
There were 111 hits to a page on the mailing list unrelated to this.
42 hits were version numbers in Safari, 154 were version numbers from Chrome, from the
user agent string, 22 hits were from web crawlers and bots etc.
And of course, 108 hits were attacks, and that's fairly typical.
So now looking just at the 4,700 and 8 media files, 21 of those were bots that can be
eliminated.
And 544 were head methods, not get methods.
So the head method is identical to the get method except that server must never return
a response.
And the reason people would do this is to check and see if the file has been changed or
not.
And then the rest were duplicate IP addresses.
So that leaves a total of 1,079, hold on, that's 414 less than what we were saying was
actually downloaded.
It turns out that people download the same episode several times on different days.
So when you put those back, you get 1,493.
Now that begs the question, should you count uni kits per day or just uni kits in general?
On the other hand, a single IP address might be hiding multiple downloads, for example
in a university or in a company or something like that behind a firewall.
So I hope you can see this as an exact science.
And even so at the end of the day, the simple fact is just because somebody downloads a show
does not mean that they actually listen to it.
Stats, I hate them.
We love them.
They also say that Hacker Public Radio doesn't like stats.
Well, that's where they're wrong.
It's just me that doesn't like them.
And that's because generating them is a waste of time.
There is no true figure that you're ever going to arrive at.
Producing the figures for this show has taken two weeks of my free time, but at least we
get a show.
So I'm happy about that.
In the process, I picked up two really cool tips from Libra Office Calc, which I'm going
to share with you.
However, every time this is discussed in the mail list, people really love statistics
and want Hacker Public Radio to have them.
I ended up putting them off for so long that the problem fixed itself.
And now that we're hosting the main feed on the internet archive, we get statistics
for free.
But can you trust those figures on the internet archive?
Yes.
By and large, yes.
And we can confirm this because we can pair their figures with what we get from the Hacker
Public Radio web logs.
I put a link in the show notes to how the internet archive works.
Each item has a view counter.
And by item, they mean like show all the multiple.
If you list all the multiple media types, WAV, FLAG, MP3, etc. would all be under one
media type.
So each item has a counter, sorry, all the media will be one item.
And the view counter is increased each time a user engages with a media item.
A user cannot increase the view count of a particular item per day.
So if I went over and listened to the MP3 and then the log, that's still only one interaction.
If a user downloads our views, multiple items on the same day, that's only counted as
one.
Now we're doing more or less the same thing on HPR.
The only difference is we count gets instead of interactions and we eliminate bots and
crawlers.
Presumably, the internet archive does something similar.
So in the example, episode for Linux in Los, season one, episode 57, the internet archive
reported 1,269 downloads while Hacker Public Radio reported 1,493.
So that's a difference of 224, but that's okay.
Sometimes it's more, sometimes it's less.
Now to explain the difference, let's explain what is actually happening here.
In an episode is published on Hacker Public Radio, it's added to the future RSS feed.
And that feed only ever points to media hosted locally on the HPR server.
And there's about 50 subscribers to that, give or take.
On the other hand, the main feed, now at least, comes exclusively from the internet archive.
So additionally, you're going to get discrepancies because initials played on the internet archive
are only going to be counted over there.
And initials played on the web page of the HPR website are only going to be counted on
the HPR website.
So there will always be differences on the download stats on both sides, but they're
close enough for jazz, yeah.
So the statement that the med that we clock in on average between 1,500 and 2,500 listeners
is a smidgen of an exaggeration.
The correct figures are 1,269 as the lowest and 2,240 as the highest.
syndication
Where the go astray is when they use that number and then guesstimate the listenership
to be between 5 and 10,000 listeners per episode.
And they feel justifies in using that number because I called, given the fact that quite
a few people syndicate us.
So some people might not know what the term syndication is and from Wikipedia, it says web syndication
is a form of syndication in which content is made available from one website to other
websites.
That's not very helpful.
So think of it as an old school content delivery network or content caching.
And this is how it would work in theory.
When the first client makes a request, the media would be retrieved from the hacker
public radio.
So instead of having the media pass through to their client, which would do, they would
also save a copy locally on their servers.
That way all requests for all subsequent requests for that file would be served from their
local website on the syndication on the syndicated website.
So anyone viewing the second or later versions of that media would not be registered in our
logs because we wouldn't see it.
Therefore, anything played there would not be counted in our internet stats.
Actually, for syndicated websites, there is a way where there's a HGTP response that
you can send over to say that this content has been played.
So even in a syndication, you can register with the source website that it has been played
or downloaded.
Anyway, but I was immediately suspicious when I heard this, not just because the legal
issues with hosting random media were because of the bandwidth costs involved.
So fun fact, no, no, that is not what's happening.
And you can prove this because fortunately, most popular web browsers have developer tools
that let you confirm exactly what's happening on the network.
So you can go to this right now, go to, for example, HackerPublic Radio website itself.
Actually, that would be pointless.
But go to Apple Podcasts, which is the example that we have in the show notes.
And if you press and hold down Control, Shift, and then press I, and then go to the Network
tab.
And then you press on any episode.
What you're going to see is, yeah, it does something on its own site.
Then you're going to see a call going out to HackerPublic Radio.
And we do a redirect to archive.org and then archive.org redirects it to one of the locals,
one of their mirror sites.
So basically that's what's happening.
So I checked.
I narrowed down the search using quotes around Linux in those podcasts and that returned
a more manageable 1,810 results.
And I limited myself to looking at all the pages given on Google and only looking at
the ones that had a play button.
And these are they in order of ones returned.
So the first three are not using the HackerPublic Radio feed, but they XML feed from the Linux
in those themselves.
They are pod chaser, player FM and YouTube.
All the rest of them are using the HackerPublic Radio feed.
And those are Apple podcasts, Apple addict, getpodcast, archive.org, list notes, Spotify,
G-Podder, digital podcast, podcast.de, hobby public radio, potency, and pod tail.
I also checked two other websites, Google podcasts, which is the Linux in those feed as well.
And I heard radio.
And all of them except two are using.
You can see when you open up all of them except two, you can see that they are coming directly
from the internet archive.
So these sites are not syndicating the content at all.
They're just syndicating their MSS feed.
So if you press play on that site, it will register as an item hit on the internet archive.
The one site that isn't there is Spotify, not because they're not hosting the media but
because they're obfuscating it.
And we were able to confirm this by looking at the HPR logs and you see that the Spotify
client user agent from different IP addresses requesting the same show is coming in.
Now if that was being cached, we would see only one IP address coming in and also subsequent
and we wouldn't see anything else for that media.
The only case for syndication is actually happening is on YouTube and the reason for that is
because they need to transfer the media from audio into a video format.
So that channel is the unofficial linux in those channels which is actually cool.
I left them an old to see how you're doing that because it will be really cool if we
go to officially do this for all the HPR shows.
And also they need to highlight that it's creative comments content, but that's a by-the-bye.
So they have 10 subscribers and a total of 606 views.
Now given the release above 57 episodes, 10 views is seems correct, but don't forget
that we need to subtract one from the hacker public radio site or the internet archive
because it was downloaded from there in the first place in order to be converted to video.
So therefore, the claim all between 5,000 and 10,000 listeners per episode is not correct.
Simply because there's no syndication going on to speak of.
Elephant in the room.
And now we need to address the elephant in the room.
But if I take a look at archive or if we take a look at an archive of an archive,
for the last one year and almost one year and a half,
we clock in on average between 1,500 and 1,500 listeners.
Given the fact that we have launched this podcast,
short of two and a half years ago, that's quite amazing.
A figure of 1,278 total downloads for the latest show is an amazing achievement.
Seriously, any podcast in the Linux space would be proud to have that.
What's even more amazing though is that they managed to garner 2,190 downloads for the very first show
because it's very difficult for new shows to get noticed.
It takes a time to build your audience and that can be seen with the grumpy old coders, for example.
They did an interview in HPR2388, which is Linux in laws season 1, episode 28, the grumpy old coders.
And they reported their downloads figures as, and I quote,
about 200 listeners across all episodes,
which they seem to agree was about right for podcasts of their type.
And I know regular listeners' podcasts would kind of agree with that.
That seems to be the norm.
Now, having listened back to that entire episode again,
it was clear that the guests from the grumpy old coders believe
that Hacker Public Radio is a podcast hosting platform.
One that operates like Spotify, Apple Podcasts, or Google Podcasts,
where each show has to build their own new audiences.
At this point, neither Chris nor Martin explained that Hacker Public Radio is not a podcast hosting platform,
but it is a podcast in and of itself,
one where the fixed RSS feed is used
by a rotating team of volunteers hosts.
Now, the Linux in laws may well believe that Hacker Public Radio
is a podcast hosting platform,
and that all the traffic is driven by their Linux in laws RSS feed coming from their own website.
Show me the stats.
So are the listeners to the Linux in laws podcast just Linux in laws listeners?
Or are they actually Hacker Public Radio listeners?
So let's compare the download numbers for the Linux in laws episodes
to the download numbers of the Hacker Public Radio episodes
that were released in the previous week to that.
We're going to look at their first episode,
which picked up 2,190 downloads in total since its release.
But on the first year of release,
it was downloaded 9,998 times.
So that's not bad.
And if we look at the shows,
four last, the 10 shows before that,
they downloaded 910, 940, 947, 968, 971,
you get the idea.
Their latest show had a first day figure of 753,
and the show released before that was 726, 722, 732, 774.
So the first day of release numbers for the first show was about 56 more
than that average HPR episode released around the same time.
And the additional downloads are common enough when a new host joins.
The first day release number for their latest show is five downloads above average
for the other Hacker Public Radio shows released the same week.
Now in the graph, I plotted all the Hacker Public Radio downloads that I know of,
and I highlighted the Linux in-laws ones on that.
And things you should be aware is that every single dot is there,
is not without cost.
There is a charge for storing it, and there's a charge for transferring it.
And it's provided to us entirely for our hosting provider,
and Anastota.com, and the volunteer project, the Internet Archive,
both of which have donated turbines of storage and data to use,
data transfer to us for free links,
and how you can support both of those organizations are in the show notes.
So looking at the graph, you can say that their shows are popular,
but you can say that they're any more popular than any other shows around the same time.
And what else can we derive from the chart?
Well, you can derive that if you want a plot count of something against a date,
in the Libra Office, you need to make sure that the dates are recognized as dates and a text,
and then you need to plot using a script or plot.
And thanks to AW35AWaf5A,
I think the show notes, you can group data by year and month,
so you've got a whole go of days and you want to consolidate them down by year and month.
You can create a pivot table using data pivot tables and start to edit,
with the days column in the raw field.
So you drag the day column into the raw fields and the sum into the data field.
And then close that, you click any cell in the pivot table,
date usually in the first column,
and you go to data, group and outline group,
and then in the section group, by select intervals,
you click both month and years.
And then you can plot those as summaries.
It's really, really quite nice.
Accurate download numbers.
We can actually determine which downloads drive from the Linux and Rows brand
and those from the HPR community.
And this is due to the fact that the Linux in-laws or SS feed
includes shows soon after they're published to the internet archive.
While the hacker public radio,
we release shows on a per schedule basis
and they only get released on their release date.
So as explained earlier, the main HPR or SS feed
will never release shows that are scheduled for a future release.
While the hacker public radio or future feed
only ever serves shows from the hacker public radio website itself.
And you can check this yourself.
If you look at the Linux in-laws feed from their own web page,
you see that they use hacker public radio.org for slash apps
and then the file that they want.
Whereas the hacker public radio of future feeds
use hacker public radio.org for slash local.
So therefore, if you go to the download statistics
for Linux in-laws shows on the internet archive link
in the show notes,
shows that are listed before the hacker public radio day
of release can only have come from the Linux in-laws feed.
And the link is screenshots in the show notes.
So for the four future shows,
they have 18 downloads,
112, 115, and 112 downloads respectively.
So on average, that puts them around 98 downloads per show.
So we can say that together with their YouTube subscribers,
their show has 107 downloads
before hacker public radio subscribers joining the party.
Now, we said the Grumpy old court has said
that they had about 200 listeners,
but they had a caveat that that was spread across all the episodes.
So it's not a per episode counts.
So they seem to match.
So is that the final answer?
Fun fact, no.
Because they do get between 1,269 and 2,240 listeners
per show.
So many hacker public radio subscribers
listen to their episode.
A number of those subscribers would also
to the Linux in-laws,
but don't because they're already getting the shows
via the hacker public radio podcast.
On the other hand,
we don't see that these shows consistently
get 107 more downloads than other shows.
So you could argue that some hacker public radio subscribers
don't listen to them,
and so would not subscribe.
Pick a number, any number, between 18 and 271 billion.
I still maintain the processing log files,
filtering them out, figuring out what's happening
is a complete waste of time.
You never get a clear answer,
and the answers can be manipulated
to get whatever results you want.
And we don't have advertisers.
We don't need to reduce numbers
to make advertisers feel better
that they're hitting their target downloads figures.
In theory, hosts may find it valuable to see
which shows is most popular and focus on those,
but in practice, there's so much variability
that nothing can be derived from the figures.
All the information that I want to know can be plotted.
How many people actually listen to the show,
and how many people were held by it?
That stuff you can't get from statistics.
The only way you're gonna get that
is if people leave feedback.
And when they do, they turn from being listeners
into community members.
Summary.
All right, closing off.
I want to explain that the purpose of the show
was not to criticize the links in those far from it.
This was intended to correct information provided by them.
They bring a wide and varied selection of content,
Tiger Public Radio, and it's very welcome
and it's indeed very popular.
Their numbers, and indeed your numbers,
if you become a HDR host,
are very impressive in their own right.
Each day, your show will be heard by as many people
as can squeeze into the Janssen room.
It had foster them.
For those who haven't been fostered them,
that's about two of those big double liquor air buses.
And every month, we have around 33 and a half
thousand downloads.
And again, to put that into perspective,
that's about 40 of those huge airplanes.
But remember, the key takeaway from this show is
who should get credit for hosting our shows?
Sinticated websites are essentially monetizing
HPR content.
They're not marrying the media,
and we're absolutely fine with that.
That's because our shows are released
under a Creative Commons attribution,
share live, 3.0 imported license.
There is absolutely no requirement or obligation
to share the spoils with us.
Another key takeaway is that our hosting
is entirely free of charge to us.
So while the podcast hosting platforms
actually host a whopping 605 kilobytes,
our hosting providers announce host.com
and the volunteer project at the internet archive
donates terabytes of storage for us to use for free.
Not just that, but also the shoulder,
the huge cost of transferring data
through expensive carrier backbone infrastructure.
The people to thank our own Josh Knapp
from an honest host.com who provides
the Hercopublic Radio website.
And the internet archive, who are a digital library,
whose state of mission is universal access
to all knowledge.
And they provide hosting for the media.
Links to how you can support
those very worthwhile projects are in the show notes.
That's it.
Tune in again tomorrow for another exciting episode
of Hercopublic Radio.
You have been listening to Hercopublic Radio
at HercopublicRadio.org.
Today's show was contributed
by a HBR listener like yourself.
If you ever thought of recording podcasts,
you click on our contribute link
to find out how easy it really is.
Hosting for HBR has been kindly provided by
an honest host.com, the internet archive, and our sync.net.
On the Sadois stages, today's show is released
under Creative Commons,
Attribution 4.0 International License.