5 ms·
That assumes that the only relevant article in the long term is the executable code. But I don't think that's true at all. The Lyons book on Unix, and several
by dwheeler 7y ago
That assumes that the only relevant article in the long term is the executable code. But I don't think that's true at all. The Lyons book on Unix, and several books about the Linux kernel, have shown that the source code also matters. If you can't correctly interpret the source code, the source code is far less useful. So if source code is also important (and it is), then having a publicly-known way to process the source is also important.
- kragen 7y agoThank you for saying this, David. I wanted to say it but I had already lost patience with the discussion. BTW, speaking of the source-code archival problem and also your work on the Karger–Thompson attack, have you been following the work on Mescc?
- dwheeler 7y agoI've been semi-following the bootstrappable builds stuff in general - that work is already listed on my page https://dwheeler.com/trusting-trust https://dwheeler.com/trusting-trust Mescc is obviously an example. Years ago Bellard demonstrated a bootstrap using tcc to compile the Linux kernel (which was pretty amazing). I think the approaches could work together well. Bootstrappable builds reduce the size of the binary you have to trust, and DDC can act as a check on a binary. Thanks for asking!