| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: Index compression vs. table compression
Tony Andrews wrote:
> Howard J. Rogers wrote:
>
>>Tony Andrews wrote: >> >>>YOU said "lookup tables". >> >>In a part of the thread that Tom did *not* quote, and to a completely
>>different person from the part of the thread which he *did* quote!!
I intended to say it, precisely as the Oracle documentation itself says it. Yes, it's a gross simplification. But to explain why they ever invented CACHE/NOCACHE, it's an acceptable one.
> Fair enough. Perhaps if somewhere you had said
> "sorry, I didn't mean to say lookup table - my mistake!" this whole
> debate wouldn't have lasted so long ;)
I am indeed sorry, that despite my very clearly having said that I did NOT say 'lookup tables' in the sentence which Tom Kyte has discussed at length, you won't accept the point. Daniel and Richard not accepting it, I can understand, since they have agendas going back years. You not accepting it, I don't understand. Read the words I wrote. Don't add any. And then deduce the meaning from them. This is usually how language works.
In my discussion with Rick, I said what Oracle itself says. (Viz: "The CACHE hint specifies that the blocks retrieved for the table are placed at the most recently used end of the LRU list in the buffer cache when a full table scan is performed. This option is useful for small lookup tables.") That is me therefore in training mode: simplifying and taking the 'party line' to get a point across (in this case, to explain what CACHE/NOCACHE were invented for). Yes, it's complete bollocks, and yes I could have said something much more like "'there is a set of "hot" buffers at the MRU end of the list and new buffers are introduced just below them. A touch count is maintained in the buffer header for each buffer which counts the number of times that the buffer has been used since its LRU status was last changed. Buffers with a touch count of 2 or more are eligible to be moved to the hot part of the LRU list'". But to have done so at that point in the conversation with Rick would have been complete overkill in my opinion. That is why I have said Jonathan's intervention was not helpful, not that it was technically wrong. It wasn't. But it was out of place. But whatever: that is one thing that was going on and being discussed.
In my brief discussion with Jonathan, however, I said something else entirely (namely, that small tables would be likely read via a FTS, even with single key lookups). I'm indeed equally sorry if that basic distinction escapes you. And I went on, in another post, to *prove* that can be the case.
I would indeed apologise if I had anything to apologise for. Meanwhile, everyman and his dog seems to think they know better than me what I *meant* to say, even though what I actually (and very carefully) said is there in black and white.
So I don't have to say "sorry, I didn't mean to say lookup table" because I DIDN'T SAY 'lookup table' IN THE FIRST PLACE!!
I said it in discussion with Rick. So does Oracle in the Concepts guide, because they are trying to get a point across. I did NOT say it in discussion with Jonathan, because he doesn't need simplified points explained to him.
As I've posted here before: words have meanings. And if I'd meant to say 'lookup tables' in the sentence with Jonathan that Tom has picked up on, rest assured I would have said 'lookup tables'. But I didn't, so I must have been referring to something else. And my subsequent post to Tom, *before* he started talking about what he assumed I'd meant, explained in totally unambiguous terms what *I* meant by 'single key lookups': a search of EMP for an EMPNO. Would you call EMP a lookup table?
No, I didn't think so.
Regards
HJR
Received on Sun Jan 23 2005 - 17:53:28 CST
 |
 |