Build1 distinct publisher2 min readUpdated
A managed C# node and a VL1-only ESP32 build show what drops out when you skip virtual Ethernet: no IP stack, no network membership, and no delivery guarantee either.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Strip VL2 away and the remaining service is narrow: a datagram sent to a cryptographic identity, with path selection and NAT traversal happening below it [3][10]. No address assignment, no ARP, no routing table, because there is nothing to be a member of [12]. On a sensor that needs to push a few hundred bytes to one server and hear an answer, that removes most of what an IP stack was there for. The same split pays differently upstream: ZtSharp exists so a .NET process can speak the protocol without libzt, P/Invoke, or per-platform native binaries to maintain [6], and it carries its own ARP, IPv4, IPv6, UDP and TCP above the virtual Ethernet layer to do it [8].
The bill lands in firmware. USER_MESSAGE offers no delivery guarantee, and the author's own list of what an application must add runs to five items: delivery confirmation, sequence numbers, retransmission, duplicate detection and ACK semantics [13][16]. The ESP32 test implements the cheapest possible version of that, a reply to each `hello from esp32:N` counted as an acknowledgement by the application rather than by the transport [14]. For a demo counter that is fine, and the author is explicit that small telemetry and control traffic is often served well enough [15]. A fleet where a missed command matters needs a retry policy and sequence state that survives a reboot, and neither arrives with the protocol.
The quieter loss is admission control. Network configuration is a VL2 function, and in ZtSharp the controller sits inside the node itself [4][7]. A VL1-only device joins no network, so no controller ever authorizes it and none can revoke it [17]. Acceptance moves into application code, which decides what to do with a sender address and a 64-bit typeId [10][11]. That is a reasonable trade for telemetry, provided somebody notices that the revocation story has changed hands.
The one interoperability result on offer, according to the post, sits at VL2 [9], and in the ESP32 exchange the counterparty is identified only as a remote peer [18]. The half of the design that is new is the half with less evidence behind it. Nor does the material carry a flash figure, a RAM figure, or a peer count for the VL1-only ESP32 build; the writeup as supplied stops mid-sentence at the server-side diagram [19][20]. The architecture reads as sound and the sizing is simply unknown.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
The author put the ESPB interpreter project on hold because interest in it was limited, and approached the larger problem through ZeroTier instead.
The work resulted in three related projects, all centred on ZeroTier.
VL1 (Virtual Layer 1) is ZeroTier's cryptographically addressed P2P layer, providing identity, peer discovery, cryptography, path management, NAT traversal, and the exchange of control and application messages.
VL2 (Virtual Layer 2) provides virtual Ethernet on top of VL1: virtual networks, Ethernet frames, network configuration, and IP connectivity.
ZtSharp is a fully managed implementation of a ZeroTier node in C#/.NET; its protocol core is based on the ZeroTier One node implementation, while the networking layer, APIs, storage and managed infrastructure are written in C#.
The stated goal of ZtSharp is to make ZeroTier usable from managed .NET applications without requiring libzt, P/Invoke, or platform-specific native bindings.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single self-authored post, architecture described but nothing measured
All claims rest on one dev.to article written by the projects' own author. The technical description is internally coherent and unusually explicit about limitations, which supports the design claims, but there is no independent verification, no code review, no reproducible test detail and no numeric evidence: no flash/RAM footprint, throughput, latency or peer-count data, and the supplied text breaks off mid-sentence. Interop and ESP32 results are single-sentence self-reports.
Author-only usage; no external users evidenced
The only observed usage events are the author's own: a project announcement, a self-run libzt interop test, and a self-run ESP32 USER_MESSAGE exchange with an unnamed peer. No downloads, package or repository metrics, third-party deployments, fleet sizes or external contributors are disclosed in the supplied material, so adoption registers barely above zero on evidenced activity alone.
Mildly overstated: broad capability lists, thin verification
The author is notably disciplined about limits, stating outright that USER_MESSAGE gives no delivery guarantee and that five reliability mechanisms fall to the application, which pulls the framing toward alignment. The overstatement that remains is scope versus proof: a long feature inventory for ZtSharp including a full managed TCP/IP stack, an embedded controller, relay and multipath/bonding, plus a working ESP32 VL1 profile, all backed by one self-reported interop sentence and no footprint or performance numbers. That leaves the capability narrative modestly ahead of the demonstrated evidence rather than dramatically inflated.
Author showcasing own unreleased-metric projects
The single source is a self-promotional developer post: the author describes projects he built, frames them against the incumbent libzt approach, and is the sole party reporting the validating tests. The post also opens by explaining a pivot away from a prior project that drew limited interest, which gives an incentive to attract attention to the new work. No commercial relationship, sponsorship, employment or vendor affiliation is disclosed, and no monetization, pricing or licensing is proposed, which bounds the distortion below the level typical of vendor marketing.
Design claims credible, outcomes unconfirmed
Confidence is limited by single-sourcing and author self-report. The descriptive layer claims (VL1 versus VL2 responsibilities, USER_MESSAGE fields, what a VL1-only profile omits) are self-consistent and check out against how the post itself defines the layers, so they carry moderate confidence. Everything about working behaviour, performance and durability of the three projects is unverified, and the truncated text removes part of the intended argument, so the overall assessment stays low.
build
The Arduino ceiling has an address: 20 kHz sampling next to a live WiFi stack1 distinct publisher
build
A $10 mesh robot moves the classroom bottleneck from the purchase order to the radio1 distinct publisher
product
Pinned toolchain images were the easy part; scoping the agent is the new build problem1 distinct publisher
build
COBOL to Java now has a price tag: £200k to £800k, and the risk is your decimals1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 23, 2026