Re: Explain-lite (Explain discussion)

At 12:21 17-11-00 +0000, Rob Bull wrote:

>but you still have the Explain search/retreival database model to create -
>in most cases where  Z39.50 servers have only a single database this
>model is a significant, complex overhead that is time consuming to add
>after an Init. Explain lite is aimed at the cases where servers have
>between 1 and several databases.

That would be *all* cases, Rob, but as Bob says, clearly Explain lite gets 
increasingly heavy as the number of databases increase.

> > And XML is going to be different why??  TO me that is no different than a
> > change to the ASN.1 definiton.
>Because working in ASN.1 is not what our partners want to do  - people
>want to work in XML - they dont want to work in ASN.1. XML by comparison
>is easy, readable, flexible, has a wide range of supportting tools and it
>fits into corporate business and training models.

This argument never made sense to me. Our partners also don't like to work 
in ASN.1 - often that is why they partner with folks like us in the first 
place. So we go to great lengths to provide tools that allows them to work 
in XML or *whatever*, while still enjoying the benefits of an 
interoperable, standard protocol on the wire. Mapping is the key. To 
attempt to adapt the interoperable standard to whatever is the happening 
buzzword at the moment (XML, CORBA, SOAP, MIME headers, etc.) is a risky 
undertaking. You've got to provide *tools* to allow people to work easily 
and comfortably in a standards-based world based on the type of environment 
they come from. But you can't have cross-platform interoperability without 
compromise.

But hey, there are exceptions. Few would argue that the Z39.50 community 
made a good business choice when it hopped off the OSI stack and onto 
TCP/IP.    :-D    Difference is, there's probably never been more than one 
(1) or two (2) OSI-based Z39.50/SR targets in operation at any given time, 
so the jump was an easy one to make. We have a much bigger installed base 
today, and we are finally at a point where people are building and 
deploying real, time-saving, hard-working applications based on Z39.50. We 
have some serious momentum going here.

The problem with Explain is *not* the encoding - it is the over-engineered 
semantics.

>It is a negotiated model - the content is what our cliehnts need, and as
>Sebastian mentioned, is date stamped.
>
>But, I would gauge that the amount of data to transfer regarding a 300
>database system in the explain model with search/present requests and
>responses is going to be very significant as well.

Difference is, using full Explain, the client can ask for a pretty compact 
list of databases using Explain, and then only download the detailed info 
when and if it is needed - and each piece can be cached separately. It 
scales very nicely.

Does ONE-2's Explain-lite support a proxy model, where a single server 
might supply info about a bunch of other servers? This seems to me to be a 
potential key feature to get Explain rolling, once we agree what we want 
from it.

--Sebastian
--
Sebastian Hammer        <quinn@indexdata.dk>            Index Data ApS
Ph.: +45 3341 0100    <http://www.indexdata.dk>    Fax: +45 3341 0101

Received on Friday, 17 November 2000 07:08:49 UTC