Date:Sun, 25 May 2003 10:09:30 +0200
Reply-To:"Z39.50 Next-Generation Initiative" <[log in to unmask]>
Sender:"Z39.50 Next-Generation Initiative" <[log in to unmask]>
From:Adam Dickmeiss <[log in to unmask]>
Subject:Re: SRW/SRU and Metasearch products
Comments:To: "Z39.50 Next-Generation Initiative" <[log in to unmask]>
In-Reply-To:<[log in to unmask]>
Content-Type:text/plain; charset=us-ascii
On Sat, May 24, 2003 at 11:02:10PM -0400, LeVan,Ralph wrote:
> No, I'm busy doing other stuff.
>
> I am opposed to multi-database stuff. As I understand it, the content
> providers get hit on multiple databases by the metasearchers. They think
> the cost of multiple connections is too high and they'd rather have a more
> complex protocol that would allow them to support more traffic at a lower
> cost. Sounds to me like they should be doing classic Z39.50. They don't
> want to do http; too many connects and reconnects. They should be doing
> sessions. Seems like they ought to work out something with the
> metasearchers to keep a session open all day and run all their traffic
> through that session. (Okay, maybe a pool of sessions.)
>
> These sound like serious folks with specific requirements and a commitment
> to serious code. Make them do real z39.50.
Are you saying that Z39.50 is serious while SRW is not? Can't be:)
We've been running Z39.50 in a semi stateless way for some time
now. In essence we have a pool of Z39.50 connections shared
by many users. In this mode, the init request is protocol
negotiation and nothing else. At least one of our customers
carry user information in user information.
With SRW and HTTP keepalive the mechanics becomes equivalent.
And besides a constant factor for extra XML encoding/decoding
the two protocols ought to be equivelent.
For SRW, we already have the notation of a database and is using
the HTTP path. That has two drawbacks:
1) can only specify one resource.
2) is bound to HTTP (SOAP can operate on a varity of protocols)
-- Adam
> Ralph
>
> -----Original Message-----
> From: Annie Denenberg [mailto:[log in to unmask]]
> Sent: Saturday, May 24, 2003 1:27 PM
> To: [log in to unmask]
> Subject: Re: SRW/SRU and Metasearch products
>
>
> Sebastian - Thanks. There might be a compelling case for building
> muti-database capability
> into SRW. Then again maybe not. Ralph's insight on this would be useful
> and I hope he'll
> weigh in here, as I strongly suspect that he would be passionately opposed
> to the idea. And (I
> think) he was at the workshop?
> --Ray
>
> -----Original Message-----
> From: Z39.50 Next-Generation Initiative [mailto:[log in to unmask]]On Behalf Of
> Sebastian Hammer
> Sent: Saturday, May 24, 2003 5:15 AM
> To: [log in to unmask]
> Subject: Re: SRW/SRU and Metasearch products
>
> Hi Ray,
>
> At 19:11 23-05-2003 -0400, Ray Denenberg wrote:
> >Ok let me ask my question another way. First, is there one of the
> >documents that
> >formalizes the metasearch model? (Appologies if there is one but I can't
> find
> >it.)
>
> I could have missed it too, ut to the best of my knowledge, no such model
> has been formalised. I believe that one of the short-term goals of the NISO
> metasearch effort is to arrive at just such a formalization -- one that
> incorporates the rights management, authentication, search and business
> model requirements of the different actors. However, at present, this is
> viewed as a fairly new busines area and there's generally not much of an
> established terminology. Even the term "metasearcher" is kind of grabbed
> arbitrarily from a handful of competing terms.
>
> >I'm assuming it's an extension of the client/server model with an
> >intermediary
> >in the middle, the "metasearcher", so the model is
> >client/metasearcher/server(s), where the metasearcher is a server to the
> >end-client and acts as multiple clients to multiple end-servers. (This
> is a
>
> Yes.
>
> >model we've attempted to formalize in Z39.50 but never could.) And if so,
> >then
>
> I actually think that we have succeeded extremely well in formalising one
> of the central elements -- the search part... no-one challenges the fact
> that Z39.50 is a comprehensive approach to solving the problem (even if the
> content vendors still see it as incomplete). When Z39.50 is not at the top
> of the list is is because of a perception among some people (perhaps a
> majority, certainly a *vocal* majority) that Z39.50 is overly complex and
> mired in obsolete technology.
>
> >there is a search protocol between end-client and metasearcher, and a
> protocol
>
> That would, in the majority of cases, be HTTP/HTML. However, it's certainly
> possible to imagine that down the road, there will be a desire for a
> web-services-like formalisation of the interface/API to a metasearcher.
>
> >between metasearcher and end-servers. And these are not necessarily the
> same
> >protocol. Are there names for these in the model, like "metasearch"
> protocol
> >and "access" protocol? (I'm asking out of ignorance; surely there's a
> model
> >somewhere that I've overlooked.) For the metaseach protocol, clearly the
>
> I don't believe there is presently a hard formalisation of these terms.
> Again, this forum is less than a year old and represents a get-together of
> many varied, sometimes conflicting business interests attempting to find a
> common platform. I'd expect to see a "vocabulary of concepts" arise from
> the email/telecon work of the group over the next half year or so.
>
> >multi-database issue is important, and for the access protocol it's not.
> (Does
> >this make sense or am I missing the big picture?)
>
> If by "access protocol" you mean the protocol between metasearcher and
> end-servers (content providers), then you're wrog -- the multi-database
> issue is seen as a core one. I know this runs against the grain of our
> discussions in ZiNG, but bear in mind that the *new* players on the scene,
> the large, commercial content vendors, very frequently run very large
> numbers (hundreds or even tens of thousands) of logical databases. For some
> of these, they allow cross-searching via their own interface, for some they
> don't, for technical or other reasons. However, the metasearchers is
> capable of cross-searching *any* combination of databases on their servers,
> basically by emulating individual users against each one. So consider
> ElseVier offering 50 different searchable, logical databases within a given
> subject area. The simple-minded metasearcher is capable of launching
> individual searches against all of these 50 databases in response to a
> single search, and this creates a very noticeable load against the server
> -- potetially a 50-fold increase in server load if a large percentage of
> users access their dataases through metasearch agents.
>
> This is why, in a nutshell, there is suddenly a requirement for
> multi-database searching in the end-server access protocol (for lack of a
> better term).
>
> >Which protocol are we considering when we talk about SRW in this model?
> >One or
> >the other or both? My reading leads me to think to me it's the metasearch
>
> SRW in this context, I believe, is being considered only for access to the
> end-servers, or content providers.
>
> >protocol, so I'm inclined to think we should reconsider the database issue
> for
> >SRW (which is a different view from my earlier posting now as I've
> re-thought
> >this). I also think we should start looking at the Z39.50 dedup service,
> >since
> >significant intellectual resources went into it, and nobody has
> >implemented it,
> >since it's really a metaseach function.
>
> I agree that a discussion about dedup is interesting, and that SRW could
> potentially evolve into a useful protocol for user access to metasearch
> engines, but that's not where the NISO group is at right now.. basically
> the metasearch engines are developed by competing companies trying to stake
> out a claim in an evolving market, and there's not presently a great
> interest in standardizing access to the metasearcher. Another concern is
> that the type of application is still so new and poorly understood that I
> think the scope of that interface would be very hard to agree on.
>
> --Sebastian
> --
> Sebastian Hammer, Index Data <http://www.indexdata.dk/>
> Ph: +45 3341 0100, Fax: +45 3341 0101
--
Adam Dickmeiss mailto:[log in to unmask]http://www.indexdata.dk
Index Data T: +45 33410100 Mob.: 212 212 66