RE: Modernization and multi-member files

On Wed, 2008-07-16 at 21:10 -0700, Alan Campin wrote:

<snip>
FWIW, from the beginning I never saw the point of multi-member files,
except for source.
</snip>

I agree about the multiple members. I think that whole idea was one of the
big failures of the System 38. A really bad idea that we still live with
today. Source could have handled like QDLS library until we got the IFS.

Indeed, I remember hearing a presentation by an IBMer where she said
that multi-member files were designed explicitly for source code. It
was an unpleasant surprise that customers used them for data. Now, the
system as a whole is stuck with the situation.

I remeber a PDP-10 manual from late 60's or early 70's describing the
system's commitment to device-independent I/O: any operation which was
in any way meaningful for a device would be implemented for that device.
The natural result was the ability to put an indexed file on magnetic
tape. The manual warned quite explicitly that "possible" does not imply
"sensible". Ain't it the truth!

Cheers,
Terry.