"Bernhard Pieber" <bernhard at pieber.com> wrote:
> Two things I remember from a previous discussion about improving Celeste -
> brought up by John Maloney if I remember correctly:
> 
> 1) Why do we need special message ids? Couldn't we just use the offset into
> the messages file? That seems easier to me and would do away with the
> problem of duplicate message ids. I had this problem once and it is a nasty
> problem. The way Celeste was implemented then meant that I could only access
> one of the messages with the same id. :-( And it was very difficult to spot.
> Normally, you won't realize that something is missing. I had to do a full
> scan of the messages file to find that out. (You know how long this takes
> with a messages file of over 200MB...)

I had this problem also, with over 10000 duplicate ids.  I made some
minor changes to the "salvage and compact" command so that it detected
this and would assign unique message IDs.  This is in the mainstream
system right now, so duplicate message ID problems can now be trivially
eliminated by simply using this command.

I also updated the MessageID algorithm so that it was less likely to
allocate duplicate message IDs.  You are right though about the file
offset being another useful sorce of these, though it would require some
changes to how message IDs are handled (for instance if you edit a
message).

Mike