Timeline

Close to two years, in one table. The first part is the time before the repository, which the repository does not hold: its dates are mine, counted back from October 2026, and marked ≈ where they are approximate. The second part is the repository's own history, and every row links to its commit, page or release.

Before the repository

when	what	chapter
long before	a Sophos XG 330 in my network, running OPNsense - one computer, every port a network card	1
2024-12	the XGS 136 bought, with a two-year Sophos licence, shipped from Japan and delivered ten days later - taken apart on the first day, then reassembled	2
2024-12 to ≈ 2025-04	run as it was sold, on Sophos's own operating system, under its licence	2
≈ 2025-04	the decision, and the start of the project's eighteen months: a strong first-generation desktop whose support is heading for its end, kept alive on open source	2
≈ 2025-04 to 2025-12	eight months and a few weeks of survey, in the evenings: the XGS 136 beside the XG 330; other vendors' appliances compared to understand crypto and network coprocessors; teardowns; the sources	3
≈ 2026-01	the first XGS 3300 - a generation whose support still runs, a different coprocessor family, bought to reach the goal	4
≈ 2026-01 to 2026-09, in this order	the board's service headers; UART, JTAG and SPI adapters bought and tried - a certain failure	4
	reverse engineering from the vendors' open sources, their platform data and the coprocessors' own filesystems	5
	attempts, in stages and with many failures, at a firmware of my own for the coprocessor - and the board's many chips that complete its work	6
	the turn: the host learns the coprocessor's own protocol, and the vendor's firmware stays where it is; the two drivers written	7
2026-09-25	the repository's first commit	7

In the repository

date	what	where
2026-09-25	the XGS 136's coprocessor released from reset; the management link brought up	commits
2026-09-26	every front port its own interface	#28
2026-09-27	the gate on the front ports was a handshake - and the vendor's fast path runs	#48
2026-09-30	the XGS 3300's last hop opens: the output ring granted in the block's own unit	#143
2026-09-30	the XGS 3300's front ports as network interfaces	#144
2026-10-01	one transmit buffer per slot (29% loss to 0.0%), IPv6, the crypto engine answers; the module follows the kernel	the road, #203
2026-10-03	offload switched on - it never had been	#210
2026-10-06	the flow table and its sweep, on the XGS 3300	#255
2026-10-07	IPsec associations from strongSwan on the XGS 3300's crypto engine, both directions	#295
2026-10-08	tunnel connections in the flow table; the receive ring granted above the watermark	#305, #324
2026-10-11	the XGS 136 opened and photographed	#333
2026-10-13	an OPNsense update on the XGS 136: the module rebuilt for the new kernel, all fourteen ports back	armada.md
2026-10-14	the XGS 136: its first offloaded flow, and IPsec in hardware both ways	#388, #400
2026-10-16	the front ports assign themselves at the first boot; an image from xgs-image-builder installed on an XGS 3300; a second XGS 3300 on the bench, with a 4x10G module	#411
2026-10-17	three appliances on one bench, clear text and IPsec measured; the XGS 136's receive at 773 in and 934 out; one menu section, every string in Arabic	#419, #416, #428
2026-10-18	IPv6 flows at 9.24 and 9.27 Gbit/s on the XGS 3300; both tables the size the board states; four transmit queues on the XGS 136; the first release, 0.1.0-alpha.1	release
2026-10-18	0.1.0-alpha.1.1 the same day: the problem reporter tells crashes apart, the chipset's record of why the host last reset kept at every boot, a progress bar on the Updates page	release
2026-10-19	open connections cut by a block list and by a TLS fingerprint; the XGS 3300's coprocessor counts what it forwards, and pf and the interfaces count it too; unusual traffic reported; the coprocessor's DoS commands probed - it drops no address	#508, #511, #516, #518, #519
2026-10-19	the XGS 136 reads its coprocessor's counts into pf and the interfaces; a connection its fast path gives back soon after it was made waits with the host; each cage reports the medium its module carries; every page of the XGS menu renders again on appliances set up by the installer	#528, #532, #535, #537
2026-10-19	0.1.0-alpha.2: eleven fixes and twenty improvements gathered since 0.1.0-alpha.1.1	release
2026-10-19	the three appliances updated to 0.1.0-alpha.2 from their own Updates page, every step photographed; 0.1.0-alpha.2.1 the same evening: a build from a test branch is offered the release, a reboot only when the driver changed	guide, release

---

Home · How this project works

Discussion

The discussion of this page is held in the project's GitHub Discussions. Showing it here loads giscus from giscus.app and GitHub, and writing in it needs a GitHub account - what that means.

Show the discussion