RE: FETCH command behavior

[Mark Crispin <[e-mail]>]

With regard to Larry Osterman's messages about references to non-existant
UIDs, I have considered the matter carefully, and I propose to change the text
in 6.4.8 from:

      ...  A non-existent unique
      identifier within a message set range is ignored without any error
      message generated.

to:

      ...  A non-existent unique
      identifier is ignored without any error message generated.

or to (just to make it clear):

      ...  A non-existent unique
      identifier is ignored without any error message generated.  Thus, it
      is possible for a UID FETCH command to return OK without any data.

My implementation already conforms to this behavior.  I believe that this
behavior is more useful and easier to program than requiring that errors be
issued.

This is based on the belief that non-existant UIDs can be issued based upon
reasonable mathematical calculations for a client to use, e.g. "123:*" (when
122 was the highest known UID from a previous session) to get data on new
UIDs.  This is a perfectly normal situation, which shouldn't cause any error.

In the case of message numbers, a client never issues an unknown message
number for any reason other than a bug, so reporting an error is alright here.

I very strongly feel that errors should not happen in the course of normal
operation.  An error should mean "something is wrong", not "empty data".


If anybody has objections to this proposed change, please speak up now.