@Poul-Henning Kamp
Author of Vinyl-Cache (The FOSS project formerly known as Varnish-Cache), a lot of FreeBSD and various other stuff.
Relays (1)
- wss://relay.ditto.pub – read & write
Recent Notes
@nprofile1q...
We built the prototype for the ESO ELT Adaptive Optics Wave-Front Real Time Computer on a handful of 48-core Epyc machines, with a stock FreeBSD kernel.
10Kx6K matrix, 700 times per second, with microsecond jitter.
Worked fine.
(Then we did it again again ten years later on a single server, with a stock Linux kernel, because they had standardized on that.)
See:
https://www.eso.org/sci/meetings/2023/RTC4AO/01_04_suarez_valles.pdf
and
https://www.eso.org/sci/meetings/2023/RTC4AO/01_10_pedersen-kamp.pdf
We built the prototype for the ESO ELT Adaptive Optics Wave-Front Real Time Computer on a handful of 48-core Epyc machines, with a stock FreeBSD kernel.
10Kx6K matrix, 700 times per second, with microsecond jitter.
Worked fine.
(Then we did it again again ten years later on a single server, with a stock Linux kernel, because they had standardized on that.)
See:
https://www.eso.org/sci/meetings/2023/RTC4AO/01_04_suarez_valles.pdf
and
https://www.eso.org/sci/meetings/2023/RTC4AO/01_10_pedersen-kamp.pdf
This is utterly competent research.
It defied logic that Russia would not have satellites with GNSS/GPS jammers on them, but now we know which they are, and we can monitor them.
That doesn't change that anything which is reliant on GNSS/GPS signals is by definition unreliable, it always were and always will be.
https://arxiv.org/abs/2606.03673
It defied logic that Russia would not have satellites with GNSS/GPS jammers on them, but now we know which they are, and we can monitor them.
That doesn't change that anything which is reliant on GNSS/GPS signals is by definition unreliable, it always were and always will be.
https://arxiv.org/abs/2606.03673
Hvor er du henne i landet ?
@nprofile1q...
Outlawing things is politically very very hard.
It's much easier to tax them.
What we need is two wealth-taxes, one taxes money, the other taxes megabytes.
Outlawing things is politically very very hard.
It's much easier to tax them.
What we need is two wealth-taxes, one taxes money, the other taxes megabytes.
nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq7qes6mstpcsn6rg3w9fwnsau68sw9h9nga9zjy3htmegg27na6wsjd3n63 nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq7cw763xfuq8vzsvslxfssz0vq2a45aq...
@nprofile1q... @nprofile1q... @nprofile1q...
To give you an idea what I removed:
rsync still support all previous versions of the rsync protocol.
rsync has a lot of complexity to support operating systems nobody runs outside retro-computing.
rsync supports i18n conversions on the fly.
etc. etc. etc. etc.
A text-book example of "add-only-software".
I did find one bug, but so much shit had been layered on top, that I couldn't convince the maintainer.
2/2
To give you an idea what I removed:
rsync still support all previous versions of the rsync protocol.
rsync has a lot of complexity to support operating systems nobody runs outside retro-computing.
rsync supports i18n conversions on the fly.
etc. etc. etc. etc.
A text-book example of "add-only-software".
I did find one bug, but so much shit had been layered on top, that I couldn't convince the maintainer.
2/2
1
nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq04e6xt23sdp5z9nlfv9cukey28ahtzjsrewuwjqv9v7y89tjqj4q3jwuv3 nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq7cw763xfuq8vzsvslxfssz0vq2a45aq...
@nprofile1q... @nprofile1q... @nprofile1q...
The reason I spent part of my vacation on the rsync source code was to see if it could be salvaged to something I could trust.
I ended up removing approx 60% of the source code, mostly by removing what was IMO wrong or pointless features.
I wasn't quite comfortable with the result, but getting closer, but I also did not want to become the new rsync maintainer, so I decided to terminate the experiment.
1/2
The reason I spent part of my vacation on the rsync source code was to see if it could be salvaged to something I could trust.
I ended up removing approx 60% of the source code, mostly by removing what was IMO wrong or pointless features.
I wasn't quite comfortable with the result, but getting closer, but I also did not want to become the new rsync maintainer, so I decided to terminate the experiment.
1/2
1
This is nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq7cw763xfuq8vzsvslxfssz0vq2a45aqf7g4fv2h3n6nh4xtvrmyqpktz6j being asked to revert to the pre-vibe rsync in Debiam.
nostr:nprofile1qy2hwu...
@nprofile1q... @nprofile1q... @nprofile1q...
I spent a couple of weeks with the rsync source code a couple of years back, and I have have deeply distrusted it ever since.
I spent a couple of weeks with the rsync source code a couple of years back, and I have have deeply distrusted it ever since.
1
@nprofile1q...
That's been a bit of an argument in Datamuseum.dk too, in particular of automated, non-reviewed OCR interpretations should be included in the archived PDF files.
I maintain that OCR is an interpretation of the artifact and therefore: No.
That's been a bit of an argument in Datamuseum.dk too, in particular of automated, non-reviewed OCR interpretations should be included in the archived PDF files.
I maintain that OCR is an interpretation of the artifact and therefore: No.
@nprofile1q... @nprofile1q...
I was actually surprised how easy phkmalloc was and even more surprised, by how much bogosity it exposed with those weak checks.
(See slide 20 here https://papers.freebsd.org/1998/phk-malloc.files/phk-malloc-slides.pdf)
That experience is a major reason why 10-ish percent of all source lines in Vinyl Cache are asserts, and I have never regretted that one second.
I was actually surprised how easy phkmalloc was and even more surprised, by how much bogosity it exposed with those weak checks.
(See slide 20 here https://papers.freebsd.org/1998/phk-malloc.files/phk-malloc-slides.pdf)
That experience is a major reason why 10-ish percent of all source lines in Vinyl Cache are asserts, and I have never regretted that one second.