[OT] Secure storage

[Tom Anderson]

That's certainly the current view, but as you said, it didn't happen
before Baby used a Williamson tube (modified CRT) to store program and
data.

However, there were a number of computers sold before that.

Lots of things sold as computers . I suppose i'm saying that the
definition of computer has changed, and that we wouldn't consider Colossus
or initial-ENIAC computers today. I certainly accept that the name was
used differently in the past.

I must admit, though, that IBM spent a lot of time (and made a lot of
money) selling stored-program computers that emulated those early
machines. Think of any small business machine (S/34, S/36, S3) with its
data files on cards and running RPG. They were alive and well until well
into the late 1970s.

True. I was reading about the IBM 1410 a while ago;

Agreed, though I can't remember whether they were programmed with a
plugboard or whether the program was one of those sliding plate memory
stacks.

ISTR it was a reel of film. I can't remember how the instructions were
encoded on the film - as holes, rather than image, i think? Hence the lack
of branches - the mechanism was only capable of reading out instructions
one by one in lockstep, because a branch would involve precisely
fast-forwarding or rewinding the film, which the hardware didn't support.
That said, i have a vague memory that there were predicated instructions,
like in the ia64, so there was a limited form of conditional logic, just
not general branches. Indeed, exactly as with the ia64, this was a way of
shoehorning conditionals into a very long pipeline!

Some latter-day clever-clogs figured out a way to make a Z3
Turing-complete, but it involved splicing the film into a loop, and
repeatedly running a fixed sequence of instructions that carried out the
cycle of a Turing machine in the Z3's store. Dubious, but an excellent
illustration of Wheeler's "any problem in computer science can be solved
with another layer of indirection"!

That seems correct from the limited amount I've seen about them. I've
seen a bit about the arithmetic unit and almost nothing about the
instruction set - only that the instruction feed used essentially the
same mechanism as a Jacquard Loom.


The steckerboard and reflector disk was there on all the Whermacht's 4
disk machines, but I thought the 5 disk version only appeared quite late
in the war.

I stand corrected. I remember this too vaguely to be any use, i think.

If I understand the layout of the rebuilt Bombe, it has the possibility
of using 6 disks - at least it has six rows of 12 disks spindles on its
face. I don't remember hearing an explanation of how these mapped onto
the disks, reflector and steckerboard in an Enigma machine - just
assumed that each column represented an Enigma machine so it could run
the problem in 12-way parallel


were you ever at GCHQ? A friend who was used to say that a lot.

I don't, but i do wonder if it might be a good idea - them or MI5 or MI6,
or one of the other minor organs of the intelligence system. To work on
really sensitive stuff, you need to have 'directly vetted', aka DV,
status, and i hear from my shadowy contacts in the intelligence world
there is apparently a permanent shortage of developers with DV status, and
an almost complete absence of *good* developers with DV. So, if you can
get as far as getting DV (by starting on a less sensitive project, getting
the first level of clearance, whose name i can't remember, then
maneuvering into getting pulled into a more serious project where they'll
put you through DV), you essentially have a job for life. Possibly on
quite interesting stuff, too, although i would guess, given the
aforementioned shortage of good developers, that most of the existing
systems are even more horrific than the festering abominations one comes
across in the commercial world.

Using it was described as keying each letter of the alphabet in turn and
seeing if the same letter appeared on the output lamps.

Yes, that fits with what i remember about running the signal through both
ways.

If that worked, they put the settings into a British cypher machine that
had been rewired to emulate an Enigma machine and typed the first
sentence or so of the cyphertext into that. If good German came out,
they passed the settings and cyphertext on to the analysts and reset the
Bombe for the next problem.

Bletchley Park is well worth a visit. I had pretty high hopes of it, but
even so it was a lot more interesting than I expected, not least because
that have a good collection of working Enigma and Lorenz machines.

They have all sorts of good stuff. If there's anyone else still reading
this thread, you are urged to go! Photos from my visit are here:

http://www.flickr.com/photos/twic/tags/bletchleypark/

I'd hoped to see the Bombe running, but it wasn't run the day I was
there, however, Colossus was up and running.

Did you meet Tony Sale, who rebuilt it? He was there when i visited.
Really nice old chap, with lots of interesting stories. In fact, i'd had a
similar discussion to this a bit before going, so asked him whether
Colossus was a computer - he was very definite that it wasn't!

tom

 

[Martin Gregorie]

ISTR it was a reel of film. I can't remember how the instructions were
encoded on the film - as holes, rather than image, i think? Hence the
lack of branches - the mechanism was only capable of reading out
instructions one by one in lockstep, because a branch would involve
precisely fast-forwarding or rewinding the film, which the hardware
didn't support. That said, i have a vague memory that there were
predicated instructions, like in the ia64, so there was a limited form
of conditional logic, just not general branches. Indeed, exactly as with
the ia64, this was a way of shoehorning conditionals into a very long
pipeline!

I don't know the ia64 but, and I hate to admit it, I have written RPG3
and it was memory of preceding a command with indicators that controlled
whether it was executed or not that brought on the small diatribe about
the IBM S/3 and friends. RPG3 is an abomination - the world's only
machine-independent assembler.

Some latter-day clever-clogs figured out a way to make a Z3
Turing-complete, but it involved splicing the film into a loop, and
repeatedly running a fixed sequence of instructions that carried out the
cycle of a Turing machine in the Z3's store. Dubious, but an excellent
illustration of Wheeler's "any problem in computer science can be solved
with another layer of indirection"!

Cool.

Did you meet Tony Sale, who rebuilt it? He was there when i visited.
Really nice old chap, with lots of interesting stories. In fact, i'd had
a similar discussion to this a bit before going, so asked him whether
Colossus was a computer - he was very definite that it wasn't!

No, I didn't meet him - unfortunately.

I love the thought that Tommy Flowers designed around telephone exchange
modules and that this, together with the wholesale scrapping of valve
exchanges, were the only things that made the rebuild possible.

 

[Mike Amling]

One of the more interesting illustrations of Lew's point is Enigma, the
system used by German cryptographers during WW II. When Enigma was devised
in the late 1930s, Germany cryptographer's calculated that it was such a
difficult cypher that it would take 15,000 years to crack. They were
satisfied that this was an extremely secure cypher and had every confidence
that it would remain unbroken for a VERY long time. As it turned out, the
folks at Bletchley Park had at least partially cracked it - to the point
that they were able to read parts of the Enigma traffic - in just 9 MONTHS!
Of course, the Germans probably hadn't expected the first primitive
computers to come in handy for this problem.....

The Poles were reading Enigma traffic before Bletchley Park was founded.

--Mike Amling

 

[Arne Vajhøj]

Am 01.03.2010 21:43, schrieb Arne Vajhøj:

Actually AES I would never try to encrypt anything with known
plaintext... too many attacks go into that direction..
i.e. use CBC not ECB ...

That is wellknown.

Most people have hopefully seen the Tux pictures.

But besides this one example of a more secure Encryption than AES-256 Bit
is using AES-128 Bit ..
as recent attacks show the keylength is exploitable and makes the
attack on AES easier.
Strange result.. though be aware of it AES-128 is more secure than
AES-256 currently

I don't think there is evidence for that.

It has been shown that encryption using 11/14 of AES 256
bit is vulnerable to attacks (that if I understand correctly
requires access to plaintext samples and use of multiple
secret keys that are related in a known way) while AES 128
bit does not suffer from the same problem.

That does not prove that a full AES 256 bit is less
secure than a full AES 256 bit.

Arne