Re: Another problem with abandon

[Simon Spero <ses@unc.edu>]

Hallvard B Furuseth wrote:

> > I don't think so.  It has the effect that any definition of LDAP for a
> > protocol which doesn't guarantee order, must reserve message IDs of
> > abandoned operations until the next bind.
> >
> > Oops.  It must reserve the message IDs of abandoned operations, period.
> > The server might receive the abandon after the bind.
[The message IDs only need to be reserved for 2MSL (for the same reason 
that TCP is supposed to wait that long for TIME_WAIT blocks to expire)]

CLDAP has never been widely used, and has always had a lot of problems 
dealing with  dropped packets and the mapping to a best effort delivery 
service.  For example, if a search request generates a number of search 
responses, there's no way to detect if any of the matches are lost, 
unless all replies are required to fit in a single datagram.  This makes 
CLDAP pretty useless, since it uses the  same high overhead encodings as 
regular LDAP.  

Whoops - I just rediscovered that I was supposed to write something up 
on this eight years ago (mar 94).  

Simon