[Info-vax] HP wins Oracle Itanium case

[glen herrmannsfeldt]

ChrisQ <meru at devnull.com> wrote:
> On 08/22/12 19:52, Johnny Billquist wrote:

>> Uh... No. At the lowest level of a disk, you do *not* store a stream of
>> bytes. Where on earth did you get that from? At the lowest level, a disk
>> deals with disk *blocks*. You read/write one block at a time.
>> Blocks are typically 512 bytes, or possibly 2048 or 4096 bytes nowadays.

> Err, thanks, but I am aware of all that. It is a stream of bytes, it just
> happens to be formatted into blocks on the disk, for identification, error
> checking and recovery.

> If you read back in the thread, you'll see that what i'm really getting at
> is that a byte  stream is the most generic / lowest common denominator, from
> which all other data formats can be layered on top of.

I suppose that is true for VMS. The IBM disks from the S/360 and S/370
days would keep block boundaries through to the hardware. You could 
write a file full of one byte blocks, and the disk hardware would
write those one byte blocks onto the disk. Not just a software
simulation for the blocks, but actual blocks.

Later disks did many tricks in the hardware, but the block boundaries
were (and are) still seen at the channel interface. (I believe
similar to the way tapes work on VMS, or at least unix, where you
can write blocks of any size.)

> Since there are possibly an infinite number of data structures that 
> you might want to develop in the future, it doesn't make sense to 
> lock all that down within the file system itself...

Maybe that wasn't so obvious 50 years ago. Still, the underlying
data structures usually have records and blocks.

-- glen