---
title: roundtable_ discussion on Off-the-grid network
tags: summit2026, program, programming, summit, 議程
---
# roundtable: discussion on Off-the-grid network

> edit welcome [name=paulpengtw]

# mentality
- what will happen when most of the network capacity are gone
- some speculation: nothing connect to the internet will work

# what should we prepare to get islander interconnected within the island

## case study
- [dComms working in Ukraine](https://drive.google.com/file/d/1Kf45_w129YS3GOH7H_hPnFn-o1cOOsyC/view)

## two aspects
- user side
- server side
- Network Time Protocol server
- TLS cert
- TWCA might midigate the issue
- ccTLD (.tw)
- DNS resolvers
- CDN edges
- Package mirrors / registries: security patches may not be available when global internet is gone, attackers may exploit this
- auth
- single sign on (e.g. google, line, facebook)
- recaptcha
- mobile push notifications
- public cloud
- security update
- Can ActivityPub help? (from the perspective of g0v.social admin)
- 
- multiple layers act as foundation: DNS, physical cables, etc
- are activitypub or matrix depend on dns
- yes
- domains that are not .tw based would not be resolved when undersea cable are severed.
- e.g. g0v.social
- neet to talk with TWNIC people for record expiry
- let's encrypt let people have CA on ip address
- maybe use a failover on dns resolver that is somehow controlled by gover't(e.g. HINET)
- Network environment in Ukraine under war
- Internet may be down for days during shelling
- Internet may be (forced) routed through Russia
- place the server inside the island or outside?
- when the outage happen, is that ideal for us to put the server inside the warzone?
- depend on the threat model and what's the hypothesis
- dComms service distribution
- instances around the city
- each is localized, citizens signup within city
- each city got their own instance
- based on mesh vpn tech
- like between Kiev and Kherson, since ap doesn't do routing
- lighthouse
- [yggdrasil network](https://yggdrasil-network.github.io/)
- tinc
- city-wise communication seems feasible; but for people hosting services it's another problem to solve
- connection between two city is served
- how to reroute: be creative, utilize as much resource as possible (e.g. computer center in university)
- what's the content people is posting on dComm services?
- chatting
- family posts
- alert channels about bombing info
- recently installing PeerTube, contents:
- tech tuturials
- politicial contents
- Impression of using Matrix: slow, huge in size, has that changed?
- It depends
- server of matrix: maybe they resolved the issue if you run on a newer machine
- is IRC/XMPP🕊️ a better alternative to Matrix?
- Matrix's UX is superior to end-users
- and they got video calling
- human organization on maintaining the service
- coops/company/org ?
- trust of service deployer is everything
- is content moderation an issue
- [simplex chat](https://simplex.chat/)
- another signal alternative
- is e2ee matter under network outage?
- what if the server got control under PLA?
- "adversarial hosting" with full disk encrytion
- boot to boot encryption with yubikey

## wifi based intranet
- cost to each device is way too high
- the range between two access point (based on ARDEN, Amateur Radio Emergency Data Network) are not good enough
- even solution like meshtastic for end user is not that user friendly
- a tripod / dish / wifi access point each will be suffice for setting up a wifi intranet AP
- case [NYCMesh](https://www.nycmesh.net/)
-

## what if certain city lost the connection to upstream service
- just call or text its syssadmin
- meshvpn access or reroute/ starlink
- the whole things is to figure out to contact someone near the server keyboard and get 2nd

## Are datacenters targeted by the enemy?

- Ukraine uses sattlelite communictions early on, thus there is no datacenter to attack

## drill/test, but what are the test items