 | |
Natural keys vs Aritficial Keys
comp.databases.theory comp.databases.theory
 |
| | Thread Tools | Display Modes |
#41
 
| |||
| |||
|
Re: Natural keys vs Aritficial Keys -
05-22-2009
, 10:28 PM
|
On May 22, 9:10 pm, paul c <toledobythe... (AT) oohay (DOT) ac> wrote: Bob Badour wrote: paul c wrote: Bob Badour wrote: paul c wrote: ... Oh, just remembered another one - fixed-point decimal arithmetic! What do you need that for? To get the same answer as the lawyer with his amortization tables. Integers are integers no matter the base. Sure they are, but I was talking about decimal points. Eg., it bugs me that the most widely-used (that doesn't mean most popular) cpu 'architecture', Intel's, can't express the fraction 2/5 exactly. If expressing exact rationals is what you want, then that is trivially done using integer arithmetic - as is fixed point decimal. Hardware decimals, which essentially died with the VAX, don't help you express rationals. Have a play with these: http://gmplib.org/ (see mpq for rationals) http://docs.sympy.org/ ..... |
The point has nothing to do with rationals, some decimal fractions are
irrational. I never said a cpu should express exact values, rather it
should express the exact same values people who are accustomed to
decimal arithmetic or traditional slide-rules come up wiith. To talk
otherwise is to argue for Betamax.
#42
| |||
| |||
|
|
If expressing exact rationals is what you want, then that is trivially done using integer arithmetic - as is fixed point decimal. Hardware decimals, which essentially died with the VAX, don't help you express rationals. ... |
which stood for "Dennis' Programming Language". Probably the Beaver was
into it too.
#43
| |||
| |||
|
|
VAX died for other reasons, not that one. One I can remember is DPL which stood for "Dennis' Programming Language". Probably the Beaver was into it too. |
PCs ... old post with VAX sales sliced & diced by model, year,
US/non-US:
http://www.garlic.com/~lynn/2002f.html#0
mid-range 370s (43xx) competed with VAX in the same mid-range market and
saw similar growth and fall-off ... although 43xx actually out-sold
VAXs, in part because of large customer accounts that would order 43xx
machines in units of hundreds at a time.
the followon to the 4341 was the 4381 ... which was expecting to see
similar explosion in number of machines sold ... but by the time 4381
came out ... the mid-range market was already starting to be taken over,
as lower-end machines (workstations and PCs) were moving up (taking over
the mid-range market).
the following post
http://www.garlic.com/~lynn/2002f.html#1
quotes from a jan88 "dec professional" article about OLTP in vax cluster
.... and the overhead cost of the (vaxcluster) distributed lock manager.
one of the things I did when starting work on DLM for HA/CMP was talk to
sybase and ingres about shortcomings in vaxcluster DLM and worked on
mechanisms to minimize the overhead (as well as drastically reducing the
recovery time) misc. past posts mentioning ha/cmp
http://www.garlic.com/~lynn/subtopic.html#hacmp
--
40+yrs virtualization experience (since Jan68), online at home since Mar1970
#44
| |||
| |||
|
|
Bob Badour wrote: paul c wrote: ... Sure it can, as long as you count things in fifths or tenths. I once worked with a product that measured distances in 2032nds of an inch so that a 16th of an inch and a millimeter were each an integer multiple of the base unit. I hate to mention international standards when Celko might be lurking around to take the point off into the wild blue yonder but in this case I have say that such a system would inevitably be living in an ivory tower when it was decided by some pretty big bodies years ago that for purposes of comparison, database data exchange or not, a millimeter equals 0.03937 inches, period, full stop. |
Quote:
|
So any system that tries to handle both millimeters and inches without fixed-point decimal hardware will need to include elaborate, intricate software algorithms to do elementary arithmetic. To me, this is totally stupid but is perhaps another example of your point that regression is more present than progress. The countless hours IEEE has spent on floating-point binary amazes me, the only explanation I can think of is that humans are more comfortable studying what they are familiar with not what they aren't, which seems crazy, it's only the occasional human who has the temerity to study what he doesn't know. |
time, which is basically what floating-point binary is.
#45
| |||
| |||
|
|
toby wrote: On May 22, 9:10 pm, paul c <toledobythe... (AT) oohay (DOT) ac> wrote: Bob Badour wrote: paul c wrote: Bob Badour wrote: paul c wrote: ... Oh, just remembered another one - fixed-point decimal arithmetic! What do you need that for? To get the same answer as the lawyer with his amortization tables. Integers are integers no matter the base. Sure they are, but I was talking about decimal points. Eg., it bugs me that the most widely-used (that doesn't mean most popular) cpu 'architecture', Intel's, can't express the fraction 2/5 exactly. If expressing exact rationals is what you want, then that is trivially done using integer arithmetic - as is fixed point decimal. Hardware decimals, which essentially died with the VAX, don't help you express rationals. Have a play with these: http://gmplib.org/ (see mpq for rationals) http://docs.sympy.org/ .... The point has nothing to do with rationals, some decimal fractions are irrational. I never said a cpu should express exact values, rather it should express the exact same values people who are accustomed to decimal arithmetic or traditional slide-rules come up wiith. To talk otherwise is to argue for Betamax. |
which people using slide rules have tons of experience doing.
|
« Previous Thread
|
Next Thread »
| Thread Tools | |
| Display Modes | |
Linear Mode
| |
Powered by vBulletin Version 3.5.3
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.