Figures 10.1 and 10.2 should not display the etype, and the caption of
Figure 10.1
should become: "Filetype". I also noticed a misspelling in Figure 101.
which should
be corrected in any case: "accesible data".
Page 5/Line 18: should read "updated relatively ...".
Page 5/Line 40: same.
Page 5/Lines 26-27 and Page 6/Lines 3-4: The definition of the end of view
is not
very clear to me. In addition, on Page 5/Line 27, etype should become
filetype, and
on Page 6/Line 4, "fileetype" is misspelled.
The adjective "accessible" in "following the last whole filetype
accessible" is somehow
ambiguous to me. What does it mean exactly ? Does it mean "previously
written by me
(as a process)", "located at an offset in the file such that the upper
bound of the filetype
is located at an offset which is less than or equal to the offset of the
last byte ever
written to the file", or something else ?
I think we need to be very precise in this definition. Yesterday, I sent to
Bill Nitzberg some
figures, trying to better define the end of view, in the presence of etypes
(message
appended below).
More generally, whether or not we decide to remove etypes, I think data
representation
should be defined in section 101.1., before view, and view should mention
that the data
representation is a component of the view. In section 10.3, the first
sentence of the
MPI_FILE_SET_VIEW semantics description should also contain something
similar to:
"the data representation (predefined or previously registered - see section
...) datarep
will be used when accessing file data."
Jean-Pierre