Re: Question on Structuring Product Attributes
Date: Tue, 18 Oct 2011 18:10:43 -0700 (PDT)
Message-ID: <ce7d6363-23ec-4ba3-aedd-a029d02f7de6_at_27g2000prq.googlegroups.com>
On Oct 17, 10:05 pm, Eric <e..._at_deptj.eu> wrote:
>
> It's sort-of fun to watch two people who don't know enough trying to
> have an argument.
It is even more fun to watch someone on the sidelines, who is not party to the argument, picking up clauses and examining them in isolation.
> What on earth is "and have them in a visible column, non-pointer, form"
> supposed to mean?
Let me spell it out for you:
- I was not posting a new thread (I would have used proper terms).
- As I detailed to Jonathan, I was not writing a fully qualified
thesis (if you are expecting that, you will find it lacking)
- I was posting a response to Celko's thread. That defines the
context.
- Celko was playing "slippery fish", to avoid dealing with specific
problems that I identified (check the whole thread)
- In an attempt to prevent the slipperiness, I used Celko's terms.
- As it progressed, I used further terms make distinctions, using his
used terms as the base (I would not have used his terms, or my-terms-
that-are-based-on-his-terms in a stand-alone article)
> All you can ever see of an index (apart from its
> effect on query performance) is the list of columns indexed and some
> vendor-dependent stuff usually about physical storage.
Excellent, you seem to understand that correctly. So your question is either rhetorical or plain absurd.
If you have an honest argument with me, just state it; if there is something you wish to clarify about one of my statements, ask a question. But this picking and gnawing on the carcasses left, after others have feasted, is for the birds.
> Just to pick on one more thing:
>
> > - each PRIMARY KEY constraint builds an index
> > - each UNIQUE constraint builds an index
>
> Not "builds", but "usually needs" :
Correcting my English is petty. If you fancy yourself an editor, send me your CV, I will send you all my posts to edit, before I post.
> - nowhere is there a requirement that primary and unique keys be
> implemented with an index
Try reading the entire statement in context. You might draw a different conclusion.
> - only one index is needed and there is nothing to prevent an SQL parser
> being smart enough to realise this
Marvellous. Identify which SQL does that (post evidence).
Until then your suggestion is possibility in the future, not a
reality, which is what Celko posted (he did not say "sometime in the
future, some SQL could ...", he posted "Here is how to do this
[now] ..."), which I am arguing. I can simply state, no SQL does that
[now]. The implication is "now", on Earth, not "ten years from now",
not "if and when the vendors wake up and read Celko's post", not on
the fourth moon of Jupiter.
> - if the parser is not smart enough there may well be a way of
> second-guessing it and still ending up with only one index.
Feel free to post the method. Otherwise your suggestion is a yet
another fantasy of could-bes and guesses. Just like Celko. Any
examination reveals that his method is based on facilities that are
not available now, as stated "do this", that the method might be
useful at some point in the future, if and when some SQL vendor
supplies the the underlying facility.
My method is based on the here and now, in any SQL (I have
specifically excluded the non-SQLs), on Earth.
> And finally...
Thanks for your opinion. I will make sure I give it the consideration
it deserves, keeping in mind your demonstrated level of capability, in
understanding technical matters, and your could-be, should-be,
attitude to addressing present day reality.
Should you wish to address any of the salient points in my post, a
piece of meat, rather than a scrap, I might give your avian opinion
>
> > ... under the covers is not relevant ...
>
> True, but you don't always seem to understand where the covers are.
Regards
Derek
Received on Wed Oct 19 2011 - 03:10:43 CEST