On Sat, 27 Jun 2026, Peter 'PMc' Much wrote:

On Sat, Jun 27, 2026 at 11:52:48AM -0700, Fred Morris wrote:
! Hopefully you're referring to BSD itself...
!
! On Sat, 27 Jun 2026, Peter 'PMc' Much wrote:
! >
! > On Fri, Jun 26, 2026 at 05:57:15PM -0700, Fred Morris wrote:
! > ! Apologies in advance, I realize you're trying to run on BSD, and I only 
run
! > ! / test on Linux so it may not work. -- FWM
! >
! > That wouldn't normally be a problem. But it doesn't fit on disk.
! > I gave up when it started to compile rust (the machine has 500m ram and
! > 10g disk)
!
! ShoDoHFlo (https://github.com/m3047/shodohflo) doesn't have any dependency
! on rust. If that's occurring on account of ShoDoHFlo I'd like to know about
! your tool / build chain and the provenance of what you downloaded.

I don't know ShoDoflo, I just tried to run Your script dnstap2json.py,
as is. So I installed python311, and then the script complained:
"ModuleNotFoundError: No module named 'dns'"

Yes, dnspython.

[...]
I found the required package in /usr/ports/dns/py-dnspython, and
the prereqs:

root@wand:/usr/ports/dns/py-dnspython # make build-depends-list
/usr/ports/devel/py-hatchling
/usr/ports/lang/python311
/usr/ports/devel/py-build
/usr/ports/devel/py-installer

root@wand:/usr/ports/dns/py-dnspython # make run-depends-list
/usr/ports/www/py-h2
/usr/ports/www/py-httpcore
/usr/ports/www/py-httpx
/usr/ports/www/py-aioquic
/usr/ports/dns/py-idna
/usr/ports/net/py-trio
/usr/ports/security/py-cryptography
/usr/ports/lang/python311

[...]

So yes, it needs rust at least for building. And I remember other
discussions also, where people complained that py-cryptography now
requires rust, so this is probably not a mistake.

Well that's news to me! I will raise it with the proper authorities, although I don't know that it will make a difference. That is enough to make me question whether or not dnspython is fit for purpose any longer.

I don't want to write a DNS library for python. :-( Maybe resurrecting old code should come back into fashion... my goal was to support LOTL to the extent possible.

Anyway, Ondrey now pushed me into a promising direction. I looked into
truss and sometimes the socket seems to open successfully, and
sometimes I see an EPERM. Seeing an error is a good thing here.
Probably at some point the named honors "-u bind" and switches UID.
This needs a bit of analysis, and is likely solveable from that..

Yes, I've run into permissions issues myself. My framestream server attempts to remove and re-create the socket with the proper owner / permissions.

Happy hunting...

--

Fred

--
Visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from 
this list.

Reply via email to