[Info-vax] HP wins Oracle Itanium case

[JF Mezei]

ChrisQ wrote:

> 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.

We are getting into semantics here, but I will disagree with the use of
"stream of bytes" for disk devices.

disk devices know of individually addressable entities which we call
"blocks" which generally contain a blob of 512 bytes of data. (Some
disks may contain larger blocks and present 512 byte blocks to the OS or
bigger blocks to the OS depending configuration).

An operating system requests individual blocks. And a block remains the
smallest addressable unit on a disk.

Since disk drives are random access devices allowing you to read block
715 just after having read block 2367, it is not a "streaming" device as
a tape drive would be.

So, when  unix application sees a very long stream of bytes, it is a
concept which is built by the operating system which converts the block
based disk into stream of bytes on behalf of the user. So in a way, Unix
has a basic RMS equivalent which provides automatic conversion from
block based disk devices to streaming bytes furnished to the application
transparently of where on disk the data it, and whether it spans
multiple blocks (contiguous or not).


Similarly, RMS converts block based disk device in to a stream of
records or stream of bytes or randomly accessed records.