The restored Autumn’s Harvest Inn identifies its software as CNet PRO 3.05d. The update archives in its file area are for 3.05c. There is no matching 3.05d archive or set of release notes in the backup.

Asking the running board for its version gives the registration string in full:

The board's console. At the Main prompt the command cnetv returns the line: $VER: CNet PRO 3.05d #30300, Phillip Beazley, Autumn's Harvest Inn.
The restored executable identifies itself as CNet PRO 3.05d.

I was a C-Net beta tester. This build came from Ken Pletzer, C-Net’s author, and it was the last one he sent me. The surviving executable gives us something more useful than the version label: we can compare it with the public 3.05c build and see exactly what changed.

The 3.05c executable archive contains nine programs. Eight match the copies in my restored board byte for byte. The ninth, bbs, is the same length in both versions: 266,156 bytes. Only five bytes differ.

One changes the c to a d in the AmigaDOS version string. The other four replace one instruction:

3.05c:  30 2D 01 92    move.w $192(a5),d0
3.05d:  70 00 70 00    moveq #0,d0 / moveq #0,d0

The original instruction reads a field from a user record. The replacement sets the result to zero, twice. Repeating it fills the same four bytes without moving the instructions that follow.

The programming header shipped with C-Net names the field at offset $192, or 402 in decimal: MailForward. The surrounding routine follows an account’s forwarding destination, checks it, loads the next account, and repeats. Immediately after the changed instruction is a branch taken when the value is zero. Setting it to zero makes the routine skip the forwarding chain.

The checks for receiving mail, an open mailbox, and the mailbox size limit remain. They apply to the addressed account. The patch stops this routine from following its forwarding setting.

This looks like an edit to the compiled binary. The file length, surrounding code, and relocation tables are unchanged, and the skipped code is still present. There is no need to infer a large unrecorded release from the extra letter in the version number.

Why disable forwarding? The comparison cannot answer that. The 3.05c release notes discuss mail bugs, so a workaround for a forwarding problem is plausible. I don’t have a 3.05d release note or correspondence that establishes the reason.

The backup also limits the date. It stamped the files when they were archived in 2001, so that timestamp cannot date the patch. The available chronology puts it after the February 1994 3.05c build and before the board went quiet in 1996, still running 3.05d. A copy of the 4.07 release sits in the Inn’s file area, dated March 1996, but the board never moved to it.

There is one more detail in the recovered data: every surviving user record has MailForward set to zero. None would have taken the forwarding path in that snapshot, with or without the patch. That doesn’t tell us whether anyone used forwarding earlier. It does mean the change would be easy to miss on this copy of the board. The version string was what gave it away.