Path: dp-news.maxwell.syr.edu!spool.maxwell.syr.edu!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!elnk-pas-nf1!newsfeed.earthlink.net!pd7cy1no!pd7cy2no!shaw.ca!pd7tw2no.POSTED!53ab2750!not-for-mail X-Trace-PostClient-IP: 24.84.208.66 From: paul c User-Agent: Thunderbird 1.5.0.4 (Windows/20060516) MIME-Version: 1.0 Newsgroups: comp.databases.theory Subject: Re: Surrogate Keys: an Implementation Issue References: <1153323860.787392.82180@75g2000cwc.googlegroups.com> <1153328021.184233.52920@i3g2000cwc.googlegroups.com> In-Reply-To: <1153328021.184233.52920@i3g2000cwc.googlegroups.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Lines: 29 Message-ID: Date: Wed, 19 Jul 2006 17:05:50 GMT NNTP-Posting-Host: 64.59.144.75 X-Complaints-To: abuse@shaw.ca X-Trace: pd7tw2no 1153328750 64.59.144.75 (Wed, 19 Jul 2006 11:05:50 MDT) NNTP-Posting-Date: Wed, 19 Jul 2006 11:05:50 MDT Organization: Shaw Residential Internet Xref: dp-news.maxwell.syr.edu comp.databases.theory:43462 Paul Mansour wrote: > paul c wrote: > >> A system-supplied CANDIDATE key seems no more a bad conclusion than a >> system-supplied date but I suspect there are readers here who would >> prefer to complicate the environment by dragging in various code >> standards or other arbitrary inventions. TTM suggests this very thing >> in its 'RM strong suggestions' and can be found online although the book >> goes into more detail and IIRC avoids using the to me risky word >> 'super'. It also doesn't pre-suppose a '32-bit' value - eg., that >> could have wrap-around problems. Maybe it's another argument for always >> using views at the application level. > > I'm not really suggesting a system-supplied key, because the sytem > would never supply it to anyone. (Except in case the user does not spec > a PK, in which case it is provided. but in a different capacity --- > this is perhaps cause for confusion.) > > On this issue, shouldn't the TTM be silent? I was under the impression > it made no attempt to discuss implentation. > If it can never be exposed, then it is hardly a key as far as the db interface is concerned and all you're asking is whether an optimization is possible using a 32-bit value to indirect some other values - if you don't think you'll run out of room, go ahead, what's it got to do with db theory? p