Re: foundations of relational theory?

In article <ww_nb.59317$Tr4.141427_at_attbi_s03>, Marshall Spight <mspight_at_dnai.com> writes
>"Anthony W. Youngman" <thewolery_at_nospam.demon.co.uk> wrote in message news:VnHGw
>dHq$Fo$Ew0I_at_thewolery.demon.co.uk...
>>
>> You saw my stats and reference to "1.05"? Given a request by the app,
>> this is the number of times a MV database has to look (on average) to
>> find the data the app requested.
>
>If I understand your 1.05 correctly, then I don't think you are
>correct in thinking that 1.0 represents the theoretical minimum.
>By way of comparison, I once took over an ODBC driver codebase,
>and I was concerned about efficiency. One metric I used was
>remote procedure calls per odbc call. On the surface, it seems
>like the limit will be 1. When I took the project over, the number
>was about 1.5. When I finished some months later, the number
>was 0.2.
>

Are you saying that if I ask the database "Get me X", it will know that my next four requests will be "Get me Y, Z, A and B as well"?
>
>> The relational theory way of saying "we won't specify the implementation
>> because we may find new ways of speeding it up" comes across as daft to
>> the MV people who say "we can't make that figure less than one, and
>> we're so close what's the point of trying to speed it up?"
>
>Given your assumptions, that makes sense, but I think your
>assumptions are invalid.
>

Bear in mind that MV thinks in terms of entities. I guess you're thinking in terms of rows. If we think in terms of the invoice, I guess you're prefetching related order details, for example. I can see how you'd shrink the number of calls there.

But to me, "X" would be the *entire* invoice, including ALL the order details. Y, Z, A and B would be further invoices, and it would be extremely hard to predict which.

I think YOUR assumptions are invalid - because you need MORE fetches, it becomes EASIER for you to predict what fetches are going to come next, because they are related. We can't predict what comes next, because where you get a predicted prefetch, we've already got that as part of X.

If you make a hundred requests at 0.2 efficiency, and we get the same *data* with ten requests at 1.05 efficiency, who is actually more efficent? :-)
>
>Marshall
>

Cheers,
Wol

-- 
Anthony W. Youngman - wol at thewolery dot demon dot co dot uk
Witches are curious by definition and inquisitive by nature. She moved in. "Let 
me through. I'm a nosey person.", she said, employing both elbows.
Maskerade : (c) 1995 Terry Pratchett