[Daniel's week] August 30, 2025

[Daniel Stenberg]

# August 30, 2025

Hey friends. End of work week means this weekly email!

## Open Source Summit Europe

Monday morning I walked on to the big tage at the Open Source Summit Europe in
Amsterdam [9] and talked for about thirteen minutes in front of a fairly large
crowd. I had fifteen minutes alotted but reached goal a little early. The
title of my presentation reads "giants, standing on the shoulders of". How it
is to maintain a well-used Open Source project in 2025.

The talk has not been released on video yet - I assume that will happen within
a few weeks, but I have made my slides available [1].

I arrived in Amsterdam on Sunday evening, did my talk on the Monday morning
and then hung out in the general areas chatting with people for several hours
until about 15:00 when I left for the airport and flew back home again. The
conference of course lasts for several more days but I decided to cut it as
short as possible to minimize work interruption. And really, to make these
appearances at different conferences every now and then it only works for me
if I do them quick like this.

My talk has been noticed in media by at least the New Stack [13] and LWN [14].
The latter is behind a paywall still and I have not read it.

## rc2

As I got back from Amsterdam I put together curl 8.16.0-rc2 [2], signed it,
uploaded it to the appropriate place and told the world about it.

We have not received any particular reports or blowback on this release
candidate by the time I write this, so: smiles all around the team.

## graphs

I polished the graphs on the dashboard [3] again. Now I added titles above
each, primarily to allow us to search for them with ctrl-f on the page.

As I had to rerun the script to view the result locally it struck me that the
github cache was downloading all the issues quite ineffectively and changed to
instead download all of them in parallel compared to one curl invoke for each.
Took it down from taking several minutes to a few seconds!

I created a new graph that shows the number of lines in the curl.1 man page
through the times [4]. When I showed this off on Mastodon it was requested
that I tone down the curl symbol in the graphs so starting tomorrow they
should appear with slightly less opacity.

## anniversaries

This week my blog URL (daniel.haxx.se/blog) celebrated 18 years since its
first post [5]. 1,527 blog posts with 3,143 comments. And 168,595 spam
comments.

Similarly this week it was 18 years ago since we first landed code in curl to
support building on and for IBM OS/400, later renamed to IBM I.

This week we celebrate seven years of shipping official curl builds for
Windows [6]. Only possibly thanks to Viktor Szakats' oversight and hard work.

## openssl

Since OpenSSL has stopped providing free updates to OpenSSL 1.x, these
versions will quickly become bad for users of the open version to use and
depend on. Therefore we will remove support for them from curl going forward
[7]. As users can buy commercial support for OpenSSL 1.x, we also offer
commercial support for OpenSSL 1.1.x in curl even after the support is dropped
from the public releases.

Also: partyly because the curl project received an invite to participate in a
panel discussion about OpenSSL forks, and partly because we already have had
this as a lingering idea, we started a wiki page documented the curl project's
view on the OpenSSL fork situation [8].

## c-ares

To simplify our code base that uses c-ares and to force users into using more
modern releases, curl will drop support for older c-ares versions going
forward [10]. In March 2026 the lowest supported c-ares version will become
1.16.0, which by then will be six years old. By dropping support for older
versions we can remove several large chunks of #ifdef'ed code in curl. The
older code is probably also mostly untested.

## Wikipedia

Last week I mentioned some stats from BBC and curl usage done on their site a
random day ("just over one million requests"). This week, Timo Tijhof [11]
brought similar data from a random day on Wikipedia. (Let me emphasize that
this is just normal traffic, not an attack or so.)

45 million requests made with curl, from 113 distinct curl releases.

Of these, 32 million use the default UA (e.g. curl CLI). The other 13 million
embed libcurl with a longer UA string containing curl (e.g. GuzzleHttp/PHP,
PycURL, UnityPlayer)

At 12 million, most are curl/7.88.1.

Fascinating. 520 Wikipedia downloads *per second* done with curl. I suppose
there are probably a number of additional ones in which curl is not mentioned
in the user-agent field so we don't know.

## slop

My AI slop talk [12] has now been watched over 16,000 times on YouTube.

## Coming up

- Wednesday: the third and final release candidate of 8.16.0 ships

## Links

[1] = https://www.slideshare.net/slideshow/giants-standing-on-the-shoulders-of-by-daniel-stenberg/282693094
[2] = https://curl.se/rc/
[3] = https://curl.se/dashboard.html
[4] = https://curl.se/dashboard1.html#curl-man-page-size
[5] = https://daniel.haxx.se/blog/2007/08/28/blogging-blogging/
[6] = https://curl.se/windows/
[7] = https://daniel.haxx.se/blog/2025/08/28/dropping-old-openssl/
[8] = https://github.com/curl/curl/wiki/OpenSSL-forks
[9] = https://events.linuxfoundation.org/open-source-summit-europe/
[10] = https://curl.se/dev/deprecate.html
[11] = https://fosstodon.org/@krinkle/115098812967821467
[12] = https://youtu.be/6n2eDcRjSsk
[13] = https://thenewstack.io/the-world-runs-20-billion-instances-of-curl-wheres-the-support/
[14] = https://lwn.net/Articles/1034966/

-- 

  / daniel.haxx.se