Text preview for : 19770830_Comments_On_Janus-1_Filing.pdf part of xerox 19770830 Comments On Janus-1 Filing xerox sdd memos_1977 19770830_Comments_On_Janus-1_Filing.pdf
Back to : 19770830_Comments_On_Janu | Home
Inter-Office Memorandum
To Wendell Shultz Date Aug. 30, 1977
From Dave Redell Location Palo Alto
Subject Comments on Janus-l filing Organization SOD/SO
XEROX
Filed on: janlisfiling.memo
Below are my comments on the Document Filing System chapter of the Janus-l Functional
Spec. Since I have not read the entire document with equal thoroughness, some of these
comments may be based on lack of understanding, but for the most part, I think they are
valid questions, and in a few cases, criticisms.
Disk volumes vs File grouping:
- The exact correspondence of a physical volume to single file drawer seems open to
question, although I suppose the user will have to think of the removable volume as
a logical unit, whether that is convenient or not.
- The notion of a "folder" as an intermediate grouping between documents and
drawers seems to have gone away. I think this is too bad; folders seem like a useful
idea, and would tend to lessen somewhat the impact of the artificial nature of the
fixed-size drawer as a logical entity.
- I don't think we should refuse to remember anything about files on offline volumes.
Certainly, such information could not be guaranteed up-to-date, but the poor user
with 87 floppy disks would probably find it helpful. and it shouldn't be very hard to
retain a snapshot of recently-seen directories (perhaps a list of the user's favorite
folders, if there are such things) This shouldn't be much of a storage problem,
especially if we have a rigid system disk.
- Are there any problems with allowing non-uniqueness of file-drawer names? This
seems like the right idea, since global uniqueness would be essentially impossible to
guarantee, but is there any impact on systems which can access several document
disks at once. (E.g. multiple floppy drives, or via the X-wire).
Selection:
- I don't understand why the type-in mode of selection is necessary. I thought the
original idea was to always point with the mouse. Isn't this sufficient? Wouldn't it
eliminate the need for uniqueness of document names?
- The single notion of Current File Selection to apply to both documents and drawers
seems a little clumsy. This seems to be motivated essentially by the desire to share
the "Show Options" button; is there perhaps a more graceful alternative?
- Perhaps the notion of multi-document selection would be useful (as in DDS). This
seems to be what is actually going on when "Delete" is applied to the entire file
Comments on Janlls-l Functional Spec: Document Filing System 2
drawer, for example (rather than deleting the drawer itself, whatever that would
mean) but deleting everything is seldom useful.
- The "Reselect" command seems perverse; why not just do this automatically. (Le.
selections would be unaffected by shrink + re-expand)
Sorting:
- The fancy manual "Move" command is OK, but I can't imagine that it will be used
heavily (probably just as well; manual sorting is not a particularly pleasant task.)
Perhaps the provision of additional automatic sorting criteria would be more useful
(e.g. I use "sort by size" in DDS all the time.) This also seems like another pl.ace
where folders might be useful: to retain groupings during a sort, if desired.
- DDS allows sorting on several criteria at once; has this been thoughtfully rejected? (I
don't find it very useful in DDS.)
Access Control:
- The effective value of "Protected Status" is a little hard to assess, since the exact
context and goals of the mechanism are not stated. Is this only to catch errors, or is
it supposed to stop some sort of malicious behavior as well? (Is "privileged user"
defined somewhere?)
- The Pilot group has been thinking about an "immutable file" feature, which would
prevent a given file from ever being modified again, thus allowing it to be physically
distributed without worrying about consistency of the copies. (e.g. for fonts, code,
etc.) Would this be a useful notion for Janus-I? If not, are there any upward-
compati bility issues?
"File" command:
- It seems a little confusing to invoke the same action in three separate ways
depending on the circumstances. Perhaps they are all necessary, but I wonder.
particularly about filing by selection within the File Drawer.
Loading/Initializing disk volumes:
- I think we should be cautious about overly automatic initialization of a disk. It isn't
clear, for example, that any user ollght to be able to re-initialize a system disk. (If
someone else did it to my disk, I'd be real mad.) Similarly. if a disk is not
recognizable as a Xerox disk, proceeding directly to a disk initialization sequence
seems hasty. Might it not be a foreign disk of some unanticipated format?
- On the other hand, I did not see anything which allowed re-initialization of an
existing (Xerox) document disk. (Is this discussed elsewhere?)
- It would appear that the Pilot file system, upon seeing a disk which was not properly
Unloaded (e.g. through a system element failure) will want to perform some sort of
reconstruction process (ala scavenger) before proceeding. Presumably this should be
mentioned, along with any other times at which an automatic (or manual?) scavenge
may be initiated.
xc: Pilot Group