Re: Message Pushing and TCP Transport

[Daniel Manley <dmanley@tucows.com>]

I would tend to agree with Peter's comments.  I haven't heard anything 
concrete that says that pushing is superior to pulling.  I prefer 
p[ou]lling because it allows the registrar to retrieve messages when 
it's good and ready.  This is also simpler for the registry.  With 
pushing, the registry system must be aware of when a registrar is logged 
in so that it can push the messages out.  That means extra work for a 
system that's already very busy trying to handle thousands of 
domain:check commands and domain registrations, and DNS updates, and....  

And for registrars that connect with many connections, the registry has 
to manage multiple connections trying to push message from the same queue.

Registries (and their developers) have enough to worry about without 
getting to asynchronous communications to the registrars.

Dan

Peter Chow wrote:

>Chris,
>
>I was at the Provreg meeting and there was at least one
>implementor who was concerned with the extra complexity
>introduced by the push mechanism.
>
>As a registrar, we would only want the push mechanism to
>be added to the protocol only if there is a clear advantage of
>it over polling.
>
>On Fri, 17 Aug 2001, Bason, Chris wrote:
>
>>Eric, per your suggestion I revisited the message thread you referred
>>to. As far as I can tell there are no comments in that particular
>>thread about polling vs. pushing transfer notifications. There certainly
>>is no consensus on this topic in the thread. Furthermore, based on the
>>WG mailing list to date, I have not seen a majority consensus in any
>>message thread that identifies pushing notifications as an acceptable
>>addition to the protocol.
>>
>>Let me just restate then what I am looking for and why. I saw Scott
>>Hollenbeck's message on the list, "WG Discussion Summary with Draft
>>Document Impact" stating the following:
>>
>>...
>>Eric's XRP presentation and subsequent discussion led to a reopening of the
>>message polling vs. message pushing discussion.  Folks in the room seemed to
>>support the idea of having both mechanisms in the base protocol, with the
>>client having the ability to select a delivery mechanism from among those
>>offered by the server.  Eric will offer up suggested text to be added to the
>>EPP draft(s).
>>...
>>
>>This statement concerns me. Let me then rephrase the question:
>>1) What justification was given that received so much support for pushing
>>transfer notifications where there was little support before?
>>2) What is advantageous about using a push mechanism for transfer
>>notification in our current registry/registrar model over
>>using a pull mechanism?
>>
>>I personally see no advantage. Although, I do see several disadvantages,
>>one of which is the unnecessary complexity that will be added to the
>>protocol. If there are no clear advantages for using a push method then
>>it should not be added to the protocol.
>>
>>Chris Bason
>>
>>
>>-----Original Message-----
>>From: Eric Brunner-Williams in Portland Maine
>>[mailto:brunner@nic-naa.net]
>>Sent: Wednesday, August 15, 2001 11:51 PM
>>To: Bason, Chris
>>Cc: 'ietf-provreg@cafax.se'; brunner@nic-naa.net
>>Subject: Re: Message Pushing and TCP Transport
>>
>>
>>>Would someone please elaborate on why this push concept
>>>is advantageous to EPP (preferably a PUSH proponent)?
>>>
>>The issue has been discussed quite a few times on the list, not just at -50
>>and -51. See the general discussion of transfer notification.
>>
>>Eric
>>
>
>--------------------------------------------------------------------
>_/_/_/ Peter Chow		Chief Technical Advisor
>_/_/_/ peter@gmo.jp		Global Media Online Inc.
>_/_/_/				System Division
>_/_/_/ ICQ: 41931890		Shibuya Cerulean Tower 11F
>_/_/_/ (tel)+81-3-5456-2687	26-1 Sakuragaoka-cho, Shibuya-ku
>_/_/_/ (fax)+81-3-5456-2740	Tokyo, Japan
>_/_/_/ http://www.gmo.jp/	150-8512
>--------------------------------------------------------------------
>
>