[Cryptography] Secure erasure

[Jerry Leichter]

> 
>> The IBM 360 was the first, by a good long margin, though actually I'm not sure when IBM first published its standard.
> 
> When the 360 came out, IBM was a fairly open company. The source for the OS was available. The System 360 Principles of Operation (PofO) was probably released in much the same maner as the similar document for the DEC VAX, with history and implementation suggestions eliminated from the public version. (I don't know this for a fact, but it seems reasonable.)
> 
> The real genius in the PofO was the way it was produced. The story goes that when some implementator asked, "How is <something> supposed to work?", the question was not answered. Instead, the relevant section of the PofO was rewritten and the questioner asked, "Now do you understand?" The result was a very precise and understandable manual about a very complex system.
With the VAX, there were two different books:  There was an assembler programming manual that told you how to write assembler code, with the entire instruction set called out and published as a looseleaf binder along with all the other manuals; and there was a VAX Standard, which described the detailed operation of the whole machine and was published as a hardcover book.

Now that I think about it, the 360 Principles of Operation were more like the Standard - and there was also a separate Assembler book.  Then again, Assembler books of that era needed to describe pseudo-operations - e.g., the 360's USING - and rather complex macro processors.

Your comment about improving any section someone found hard to understand is, to this day, how I do code reviews:  I repeatedly tell people that if, in a code review, I comment that I don't understand something, the response I want is *not* for them to come tell me how it works, or even add an explanation to the review; it's to add to or improve the comments *in the code* so that the next reader won't have a problem.

Going full old-fogie here:  The manuals produced by the larger, better companies in those days were extraordinary.  The Principles of Operation really did describe, in clear, well-written English, all the details of a 360.  IBM language programming manuals were excellent, as were DEC's.  There was a DEC RT-11 manual that I used to recommend to people as one the best operating system theory introductions out there.  People would joke about taking up a whole orange (later gray) wall with VMS documentation - but everything was there, fairly easy to find, easy to read and understand.

All this is gone,  When the hardware costs millions, users are willing to pay hundreds for a set of manuals, and you can afford to pay technical writers to create and maintain that set to the highest standards.  When the hardware costs hundreds, no one can afford to put in the effort.  This was a phenomenon that first became noticeable (to me at least) in early MS/DOS days.  I wanted to learn MS/DOS the way I'd learned many other OS's.  But ... Microsoft didn't publish programming guides.  Hell, you needed to know the BIOS calls, too - and no manufacturer published those
either.  Instead, outside writers developed books of varying quality, always somewhat out of date.

Microsoft, over the years, has actually gotten much better - though the ability to put everything on-line, readily searchable, has suppressed the demand for a well thought through, organized, broad presentation.  Those are hard to find for any system these days.

> 10-15 years later, IBM started drifting down the secrecy path. :-(
DEC never did for CPU's - the Alpha standard was public - but went down that path with the BI bus.
                                                        -- Jerry