[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [search-ws-comment]DisplayTerm // Forms versus Scan and Facet // was: ....
To quote Edward: ">>Some my say: "Performance".. Its faster to search for >>"n201012345" in my LCCN index than to find "Ralph Levan". " My take on this (in addition to what Edward says): 1. Performance is not an issue here, as <query> can still include "n201012345" if the server so chooses to handle it for performance reasons. 2. And on top of that: let's not require the client to solve the server's performance problems. As an analogy: I'm sure that for my email server it would be easier if I logged in with a number and not my email adres, but why should I care? Somebody else's problem.... Ok, let's try to wrap this up. I believe we all had plenty of opportunities to voice our opinions. Since Edward and me seem to agree about this and both of us already have a working implementation, I think a fair way to do conclude this discussion is to require Ralph to show a pseudo code example to prove his point. Please not that I'm proposing this with the best intentions; I feel this discussion is starting to take too long. Regards, Edo >>-----Oorspronkelijk bericht----- >>Van: Edward C. Zimmermann [mailto:edz@nonmonotonic.net] >>Verzonden: donderdag 21 oktober 2010 14:43 >>Aan: Edo Plantinga; LeVan,Ralph >>CC: Ray Denenberg, Library of Congress; >>search-ws-comment@lists.oasis-open.org >>Onderwerp: RE: [search-ws-comment]DisplayTerm // Forms versus >>Scan and Facet // was: .... >> >>In Ralph's example: >> >> >>On Wed, 20 Oct 2010 11:54:44 -0400, LeVan,Ralph wrote >> >>> <terms> >>> <term> >>> <actualTerm>n201012345</actualTerm> >>> <displayTerm>Ralph LeVan</actualTerm> >>> </terms> >>> </facet> >>> >>> In this example, there is no index that supports looking up names. >>> There is only an LCCN index. LCCN is worthless for display >>to the user. >>> >>> Ralph >>> >> >>I agree that the cryptic n201012345 is "worthless" for anyone >>not intentionally searching a LCCN index (knowing the >>numbers) but might also suggest that a display of "Ralph LeVan" >> >>OK.. Pretend that there are more than one "Ralph LeVan" then >>a search won't return just the LCCN = n201012345 Ralph but >>the others.. On the other hand a scan would also return a >>collection of what on the display level is a list of multiple >>"Ralph LeVan"s but with different LCCNs. >> >>Pretend that our target has ONLY a single index: an LCCN >>index. Somehow we are associating Ralph LeVan with n201012345 >>and thus it does not matter if there is some >>table/relation/index in the system that associates n201012345 >>with Ralph LeVan or there would not have been the possibility >>to return Ralph LeVan in the <DisplayTerm> associated with >>the value n201012345 as its stored in the index. For the >>implementor in this case its pretty easy: If the term does >>not meet the syntax of an LCCN then its a name. >> >>The scan of the index as names would produce a list of >>names.. as LCCNs its a list of ... LCCNs.. >> >>Again.. anything goes.. And if having a list of user names is >>not what the server wants to deliver (or accept) then it does >>not need to supply them or handle them.. >> >>Some my say: "Performance".. Its faster to search for >>"n201012345" in my LCCN index than to find "Ralph Levan". >>Maybe.. but surely the resources demanded--- CPU, I/O, >>bandwidth etc.--- I would suggest are higher in the >>DisplayTerm model than in what I've proposed, Throw in the >>architecture of our computers which tend to have for our >>application I/O bottlenecks and the DisplayTerm looks even >>more costly... >> >> >>-- >> >>Edward C. Zimmermann, NONMONOTONIC LAB >>Basis Systeme netzwerk, Munich Ges. des buergerl. Rechts >>Office Leo (R&D): >> Leopoldstrasse 53-55, D-80802 Munich, >> Federal Republic of Germany >>Telephone: Voice:= +49 (89) 385-47074 Corp.Fax:= +49 (89) >> 692-8150 >>Umsatz-St-ID: DE130492967 >> >>
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]