Hi Hugh,

--On Thursday, March 20, 2003 2:59 PM -0800 Hugh McIntyre <[EMAIL PROTECTED]> wrote:

| Strongly preferred because:
|
| *  If you do the append/expunge solution then the modified message is now
|    at the end of the mailbox with no obvious way to fix things so the user
|    sees it in the original location.  This seems like a big problem for me
|    at least.

That depends on the typical sort order that users use. If they sort by sequence number (natura sort order) then yes it is broken. If they sort by date sent/received, thread etc it will be fine. Now that IMAP has support for SORT/THREAD its not as much of an isue as it was before.

| *  Append/expunge does not match IMAP's usual delete model with repect to
|    being able to store a \Deleted flag and only have this really take
|    effect on the next expunge.  With the opportunity to undo the deletion
|    before that.  It would be better if deleting attachments would work
|    the same way, not least because:
|
| *  if I have a message with three attachments and want to delete 2 of
|    them, I would prefer not to do two appends in sequence.

Clearly you can do a single append to make all the changes you want in one go. In fact right now I have a manual process like this that I give out to our users who really want to do it, and it works fine.

| I do agree with the outline of the solution you propose though, i.e.
| go ahead and nuke the data (with a "removed' or "deleted" header) but
| claim in the bodystructure that it's still there.  And then on fetches
| either return "NO" for a fetch of that specific body part or fake out
| blank content for partial or whole-message fetches.
|
| And incidentally I am one of the people who definitely wants this feature.
|
| You'd want some way to fetch an appropriate annotation (or these headers)
| to allow showing the client that the attachment was not there.  Ideally
| at least.

I think 'Content-Disposition: removed' with a parameter to indicate when it was removed and perhaps who removed it would suffice. Annotations would work as well, but I would not want this to depend on the presence of annotations.

--
Cyrus Daboo

Reply via email to