this is a little complex to explain, and please be patient and strictly
follow debian community guidelines, as i have had strokes and now am forced
to use AIs in a symbiotic fashion. below is
written *without* their assistance.

repo first, explanation second, motto at the end:
https://gitlab.com/d12rg/wot/

about 25 years ago i envisaged debian being fully peer-distributed. not
dependent on any domain hosting in any way, the entire process of debian
development could hypothetically be done via TCP/CP using git (bundles via
Carrier Pigeon) or a true peer-to-peer git transport [not to be confused
with pushing to a server, i do mean e.g. pushing git bundles to a
bittorrent network and linking SHA256 commits directly into its Kademlia
DHT]

in addition to that, many people will have seen how badly windows .EXE
provenance goes (the windows 95 sheep screensaver wrapped in malware was my
favourite) and a plethora of linux distros are blissfully unaware of the
deeply impressive provenance, documentation, rigor, and... i run out of
superlatives - i got up to 18 hard requirements and listed how every
distribution but debian fails them:
http://lkcl.net/reports/wot/

in addition to *that* if there is a *non-debian* or
even non-distro aim to distribute, for example, the
source latex of arxiv.org peer reviewed academic
journals, and have each PDF built as a binary
package and the latex as source, where the
Citations are mapped to package dependencies
and - this is the kicker - the distribution does NOT
end up messing with /var/lib/apt etc but can be
placed into an ENTIRELY INDEPENDENT location
of the user's choice...

this is where the research project i am putting together kicks in.

the goal is to then link the entire thing together behind fully
peer-to-peer infrastructure. apt-swarm is a candidate however that does not
help with the other aspects: peer-distribution of GPG keyring signing and
Identity verification protocols, whom do you trust when the entire
provenance is peer-distributable, even to the point of allowing you to
choose whom *you* trust as ftp-masters, *and* which "Releases" - quantity
multiple - you accept.

remember that in this scenario there will NOT be one single
/etc/apt/sources.list[.d] there will be *MULTIPLE*
$(repo[nnnnnn])/etc/apt/sources.list[.d]
of *your* choice for *your* chosen purpose(s), every single one of them
fully decentralised, peer-distributable, the number of top-level Release
files is not in any way limited (and can go directly into a DHT *and* have
apt search keywords associated with it)

yet still fulfilling the hard requirements that gives debian its
near-invulnerable reputation to provenance attack.

the first stage - self-bootstrap - is already complete. the scripts can
automatically package themselves up as a debian package, install and set
themselves up under an arbitrary directory, set up a debian pool mirror
ready for actual distribution, and can then be installed AGAIN from that
mirror and re-built or installed again, in order to perform gcc-style
bootstrap self-checking (does the compiler compile itself 3x in a chain and
produce identical binaries?)

the second stage is to introduce apt-swarm and separate out packages for
the documentation and associated papers, set up reproducible builds and
tests for use of apt-mirror and other hosting options, sort out the gpg
keyring peer distribution.

a third stage is to establish a rigorous set of hostile attack vectors,
actively seeking to shred the entire system and make it fall over.

i would welcome feedback and assistance, this is a fascinating
meta-meta-project that solves a particularly complex provenance and
reproducability problem on moving target(s).

until otherwise established a mailing list is here:
https://groups.google.com/g/d12rg

one particular question i would love to hear from people: i suggested
putting the entirety of arxiv.org source and PDFs into debian archive form:
what would *you* use this for? software project management of releases,
even those from Visual Studio or Eclipse or Arduino? ERM or other Sales
Force Management of customer contracts? a type of blockchain which goes
well beyond Ethereum? what else?

the motto, if you have seen mythbusters, is to use [software] tools for
purposes for which they were never originally intended. back-to-back. and
not as a rube goldberg machine.

l.



-- 
---
geometry: without it life is pointless
the fibonacci series: easy as 1 1 2 3
we are humans and can make mistakes

Reply via email to