RE: LAST CALL: draft-ietf-asid-ldapv3-simplepaged-02.txt

[Ed Oskiewicz <ed.oskiewicz@bt-sys.bt.co.uk>]

Sorry that your comments are at the end - brain dead mail program.

Paging is not necessary for simple clients, they can just suck data from
the network as and when they feel like it. Complex clients can and
should handle multiple directory searches for themselves, 'mandating
minimal resource requirements' is just a fancy way of saying "I can't be
bothered to provide the resources for my client you should provide them
for me". What people actually want from paging is for servers to store
the results of queries for them, I regard this as intolerable as IT DOES
NOT SCALE, if you want the results, you should store them.

The spec clearly indicates that paging sessions may be abandoned due to
lack of resources thus there is no pretence of a guarantee that busy
servers can keep a set of paged results long enough to be of use for a
client. Thus clients will experience intermittent failures on busy
servers and will blame the server when the true cause is the inadequacy
of the client. Worse, the directory service as a whole will be perceived
to be unreliable when the actual reason is flaky clients.

As you hint in your final paragraph, the traditional engineering
technique for coping with a mismatch in performance between subsystems
is to insert a buffer between them, if inadequate clients want a storage
service for directory lookups then perhaps we should be looking at a
intermediate set of proxies which can store the results of directory
searches performed at full speed and trickle them out on demand. Clients
can then, if necessary, be charged for the storage service they are
using.

I repeat, the burden on servers of paging as described in this draft is
unnecessary and in my view silly.

Ed
-- 
Ed Oskiewicz, B54/81, BT Labs Martlesham Heath, Ipswich, Suffolk, UK,
IP5 3RE
ed.oskiewicz@bt-sys.bt.co.uk
Tel: +44 1473 640896	Fax: +44 1473 643545

>-----Original Message-----
>From:	Vinay Shankar [SMTP:Vinay.Shankar@worldtalk.com]
>Sent:	25 February 1998 23:04
>To:	'Ed Oskiewicz'
>Cc:	'ietf-ldapext@netscape.com'; 'ietf-asid@netscape.com'
>Subject:	RE: LAST CALL: draft-ietf-asid-ldapv3-simplepaged-02.txt
>
>With respect, I beg to differ here. The paging might be an unnecessary
>feature from certain applications' point of view but definitely not
>silly. There are applications that would like to see this feature in
>LDAP servers. The servers are expected to be able to handle the extra
>burden caused by this kind of feature as they will normally be running
>on "powerful"  systems while one would expect the client applications
>to mandate minimal resource requirements.  Also it may not be desirable
>to change the underlying network configuration for a specific LDAP
>client.
>Though I do agree that eventually we will have a LDAP protocol without
>'L' but that is only expected when a access protocol becomes a "server"
>protocol.  (LDAP servers that front-end an X.500 Directory Server can
>easily provide the paging by making use of this feature in the
>Directory Server.)
>
>Regards.
>-vinay 
>
>Vinay Shankar
>Worldtalk Corporation
>
>