ORPHANWARE October 22, 2026: Millplus V410 control-disk audit
A member of the Practical Machinist community, working with a DMG Mori DMU 80 P machining center, published a complete disk image of the machine's Millplus V410 control and asked whether the image was whole and trustworthy. Before anyone restores a control from an image, that question deserves a real answer, so we audited it.
The audit was read-only and done at the file system level, the same standard we apply before any recovery work. The 1.9 GB image was checked end to end: partition layout, both FAT volumes, every file's cluster chain, and the restore kit stored on the data volume. The image's SHA-256 hash matched the hash its publisher posted, so the file audited is provably the file that was shared.
Result: the image is complete and internally consistent. Zero corruption was found, and the restore kit on the data volume was verified present and matching its own documentation. The audit, the reasoning, and the one caveat we flagged (the kit's self-installing packages were inspected but not executed, on principle) are published in the thread itself: Practical Machinist, Millplus collaboration thread, post 211.
Why publish an audit like this on a recovery site? Because examination before trust is the whole discipline. The same read-only, verify-everything approach is what the first step of every engagement here looks like, applied to a disk image instead of a program binary, in public, where anyone can check our work.
October 2026: the lab proof set
The numbers on the homepage come from a controlled set of test engagements against real compiled targets, run through the exact pipeline used for client work: decompile, reconstruct, rebuild, then differential-test the rebuild against the original.
- 9 of 9 C and C++ targets recovered to compilable source, with zero divergence across thousands of differential fuzz cases per target.
- Targets included the sbase sort, tar, and find utilities, a C++ text statistics tool, and a deliberately hostile build compiled with aggressive optimization, link time optimization, and stripped symbols.
- bzip2 1.0.8 was recovered at 8,127 lines of source. Its recon is published in full as the sample Recon Report, so you can see the working format, not just the totals.
October 19, 2026: Delphi and Object Pascal feasibility
A Free Pascal compiled target was recovered with class and property names intact from runtime type information, and verified across 4,654 fuzz cases. The caveat is stated the same way everywhere on this site: that target was built with Free Pascal, not Embarcadero Delphi, and VCL forms applications have not been tested yet. Delphi work starts with a recon verdict on the specific binary, never with a promise. Read more in the Delphi directory entry.
What does not appear here
Client work is confidential and is never published. What appears on this page is work that is already public, like the Millplus audit, or test work against public targets, like the proof set. If your engagement can be written up afterward, that is your decision to make, in writing, after delivery.