802.11s, Gemini, ggwave, Cloud, First Principle, PHA, Preallocation
802.11s (@ Wikipedia) Devices in an 802.11s network are called Mesh Points (MPs). These devices communicate directly with each other to form a mesh topology. The network supports multi-hop communication, meaning data can travel through multiple intermediate nodes to reach its destination.
I asked Gemini (@ Wikipedia) 2.5 Pro to create a Python program that visualizes file allocation on a Btrfs system, and it worked beyond my expectations. It showed exactly what I expected - all the data is 'cosmically fragmented' if it's actively updated. - That's why insanely high IOPS performance is required from storage devices, because each IO can read just a few kilobytes at time... And of course in between reading the actual data, you'll also need to read the METADATA pointing to this data which is also very fragmented. Just wondering how badly this works with QLC drives which have large storage blocks. I'd guess it provides nice write amplification!
Gibberlink Protocol (ggwave)(@ Wikipedia) - Multi-frequency encoding with 96 distinct frequencies - For fun: Cassette Data Formats vs. GGwave - Ref: Frequency Shift Keying (FSK), Kansas City Standard (KCS) - Simple binary encoding (typically 2-level FSK)
Cloud Services - "'The cloud' is not just servers. 'Going to the cloud' could also mean locking into a forever sub-contractor" (@ berthub.eu) - Why don't we think in modern terms. Of course there must be: Fungible Cloud Service (FCS) and Non-Fungible Cloud Service (NFCS) - FCS can be deployed to many / most of providers or self-hosted trivially and practically without any changes. And NFCS are the junk many companies are pushing, which forces you to pay them forever.
First Principle (@ Wikipedia) - Yeah. I'm big fan of it . Whenever I do something, I'll try to reduce it down to the basics, what's actually necessary and how to achieve it using sane amount of resources so it works reliably avoiding bloat.
When programmers learn to clean up their mess? - Never? - "Apache Guacamole does not have a built-in mechanism to automatically prune log data or access history from the database. This means that the database will grow indefinitely unless you implement separate cleanup processes to manage the size of these tables" - If you PUT something somewhere, you should also always ask the basic data retention questions. When and how the data gets deleted. - Duh!
Most interesting, got sick'n'tired about TLS blah blah, discussion. 6 weeks, no progress, just blah... Nobody even seems to know, what they actually want. Anyway, then to the point.
TLS + PAKE as PHA (2nd version) (@ datatracker.ietf.org) - I started to investigate this concept, that would it be possible to implement. Because we've had long discussion if PHA is practically working solution. Post-Handshake Authentication (PHA) can be implemented as extension / module, even when underlying TLS stack doesn't directly support necessary authentication features / modes. I didn't even know about the RFC. I did know that PAKE wasn't going to be supported by TLSv1.3 which initially made me slightly sad. But this draft is exactly what I was going to propose / suggest. Awesome! - I'm tired of one project where six weeks have been sunk into certificate discussions, and it's still not working. Yes, I'm also aware about mTLS and TLS-PSK. - kw: certificate management, password, mutual server and client authentication. ref: CPace [I-D.draft-irtf-cfrg-cpace-13], SPAKE2 [RFC9382], OPAQUE [I-D.draft-irtf-cfrg-opaque-18] and SPAKE2+ [RFC9383].
Storage Preallocation, unfortunately many programs DO NOT optimize disk access, even if the operating systems support it. From Python OS library:
posix_fallocate, posix_fadvise options, POSIX_FADV_DONTNEED, POSIX_FADV_NORMAL, POSIX_FADV_SEQUENTIAL, POSIX_FADV_RANDOM, POSIX_FADV_NOREUSE, POSIX_FADV_WILLNEED, POSIX_FADV_DONTNEED
2026-01-18