home *** CD-ROM | disk | FTP | other *** search
-
- On Mon, 12 Jul 1993, John Gardiner Myers wrote:
-
- > I maintain that it is a mistake to expose implementation details
- > through the interface unless the user is explicitly making
- > implementation-specific requests. In particular, it is wrong to
- > expose the storage format through the default "rubber-room" namespace.
-
- > Why should the user have to know or care whether the mailing list
- > archives are stored in netnews, berkeley, or tenex format?
-
- The user shouldn't know or care, neither the format, nor the absolute
- location in a filesystem...
-
- Things I believe to be true:
-
- 1. The issue is to define a satisfactory user-visible namespace, not
- expose implementation details. But the user-visible namespace
- *might* use a syntax familiar in *some* filesystem! Even if the
- mapping to the actual filesystem organization is different.
-
- 2. It is highly desirable for the sysadmin to be able to choose
- where/how to store archive or other data.
-
- 3. The user-visible namespace *must* support hierarchy, which means
- a canonical path syntax must be defined, even though it may be relative
- to a sysadmin-defined root, or a driver-defined abstract root.
-
- 4. We are really searching for a way to select an IMAP driver via the
- name string syntax that goes across the wire. However, whatever means
- is chosen to tell a driver that "this name's for you", the
- requirement to support a hierarchical path (without using netnews dot
- notation) is still with us.
-
- -teg
-
-
-
-