Re: [ldapext] nfsv4 vs the ldap consistency model

[Andrew Findlay <andrew.findlay@skills-1st.co.uk>]

On Fri, Nov 21, 2008 at 11:57:36PM -0800, Kurt Zeilenga wrote:

> If what they are after is transactional consistency, there may be  
> existing LDAP implementations which provide this.  I recall something  
> be presented in this area at LDAPcon.

Yes - the CA product has something like this, which Hilla Reynolds
talked about in Cologne last year. There is a cost to it though, as
all writes block until the slowest replica responds. There are also
some potentially nasty problems if the network gets partitioned.

There is no magic pixie dust for this: *any* distributed replication
system has to make compromises between consistency, timely updates,
and behaviour in the face of various failure modes. The Three-Army
Problem tells us that we cannot design a perfect protocol. DNS, X.500
and LDAP have made similar choices here: favour the availability of
*something* rather than blocking everything to achieve global
consistency.

Andrew
-- 
-----------------------------------------------------------------------
|                 From Andrew Findlay, Skills 1st Ltd                 |
| Consultant in large-scale systems, networks, and directory services |
|     http://www.skills-1st.co.uk/                +44 1628 782565     |
-----------------------------------------------------------------------
_______________________________________________
Ldapext mailing list
Ldapext@ietf.org
https://www.ietf.org/mailman/listinfo/ldapext