PGP: sequencing messages

[Steve Lamb]

Wednesday, April 05, 2000, 10:18:21 AM, Trevor wrote:
> Don't you see the flaw in this?

    No.

> Suppose you were talking about a POPULAR email client, such as Outlook
> Express or Eudora. The chances of many people at the same domain using the
> client on the same day are EXTREMELY good. Then you've only got 10,000 (or
> 100,000 in the third example) possible combinations. Subscribers to att.net
> or aol.com could probably guarantee a message ID would be duplicated at this
> rate.

    I don't think so.  That is why I said a "mixture of unique and random"
elements.  The unique elements would prevent duplication on the mass scale and
the random elements on the local scale.  I will admit that TB!'s scheme is not
all that great.  In fact, I believe I've filed a bug on it.  But look at
Mutt's.  What are the chances of two people composing a message at the same
time, down to the second, /and then/ a random number also being the same?  I'm
not even sure that the 2nd half of mutt's string is completely random?

    Now, AOL also has, uhm.. 20-30 SMTP servers.  What are the chances of
those servers, which are the same, with the same algorithm, duplicating a
MSGID?

    To me, there is no difference between the two.

> All this proves, I guess, is that leaving message ID assignment up to
> email client developers runs the risk of them implementing a lame
> algorithm.

    In which case it is a bug.  Just as a lame algorithm in the server would
be a bug.

> (But I still wouldn't call it a bug if an RFC specfically says that
> the uniqueness is guaranteed by the server, not the client.)

    It does not say server.  PMMail, in this regard, is clearly the exception,
not the rule.  On a technical matter such as this I do consider that a bug.

-- 
         Steve C. Lamb         | I'm your priest, I'm your shrink, I'm your
         ICQ: 5107343          | main connection to the switchboard of souls.
-------------------------------+---------------------------------------------