It couldn't be because perl is installed by default on all of the target platforms. Practicality trumps conference talks when there's work to be done, even in the government.
Whereas, the NSA's project failed initially because the team couldn't design a security kernel that had great security and acceptable performance. Told NSA they'd have to pick one. Schell told NSA he knew a guy with a design, GEMSOS, with both properties. NSA reluctantly used GEMSOS in BLACKER. The first, highly-secure VPN w/ general-purpose kernel was born. Who knows what the deployment or usability side of it was, though. Classification rules kept them from publishing on it for a decade or so where it then got paywalled. Classification is probably why Larry Walls didn't say much about BLACKER when describing its history. At least ones I read.
A few points of note: it's rather weird to call BLACKER a "VPN"; it's likely much broader than this (it's a network, crypto suite, secure kernel, system architecture, etc), and yet encompasses a very different goal. In fact, the degree to which it originates out of secure kernel research is, we argue in our paper, somewhat unclear, and perhaps this is only a small part of the equation.
If anyone has any additional information about these early architectures, I would love to speak with you, contact me at http://iqdupont.com.
http://cahighways.org/wordpress/?p=5460
Yeah, BLACKER did quite a bit. It was a network MLS component like many others that came after it. GEMSOS, Boeing SNS Server, and DiamondTek LAN are examples. Modern variant would be an Octeon 2 or 3 PCI card with similarly secure software.
Too much to explain, though. BLACKER's main purpose was securing the connection between dumb terminals and things they connected to. It used crypto, MLS, and TEMPEST-style hardware. It was a network device as well. A VPN is closest term for modern audience to convey its main goals.