29 Aug
2006
29 Aug
'06
9:34 p.m.
Don wrote:
Michael B. Brutman wrote:
The applications use OS services to create the objects. There are APIs
for creating all of the different types of objects on the system, and
the OS handles the tagging.
Keep in mind that OS/400 (or whatever the current name is) has a fairly
limited set of objects. The job queues, user profiles, subsystem
descriptions, database tables, etc. are all 'objects'. Things like
generic files that are not interesting to the machine but might be
interesting to client machines that connect to OS/400 get dumped in
generic 'stream files' in a hierarchical filesystem.
The OS enforces correct usage of these objects. You can read or access
a database table as a different type of object. The OS implements this
by storing meta data in each object and managing object types in a
centralized manner.
Mike
I don't know if anybody pointed it out already but the IBM S/38,
AS/400 and derivatives have strongly typed 'objects', not files. It
doesn't matter what you name a database table, an Ethernet line
description, or a user profile - the OS knows what those things are
and will not interpret them differently if you copy or rename them.
Does the *application* (that created the object) tag the file?
Or, must the OS be made aware of the file's type (by inspection,
etc.)?
If the former, how do competing applications "register" their
ability to handle a particular "type"? (e.g., the example
from a previous post of wanting to "open" JPG's with different
applications -- as a function of the JPG itself)
As a concession to hierarchical filesystems like
the ones used by Unix
there is a 'file system' where the OS refuses to interpret what the
files are and just considers them streams of bytes.