Shameless Promotion Re: [ic] Slow searches (fwd)
Wed, 28 Feb 2001 14:02:15 -0500 (EST)
Mike Heins wrote:
> If it is categories, you need to use the dl=string method of binary
> searching, which will bring search times under a second.
> If it is full-text, you need Glimpse, which will do much the same thing
> for word-based full-text searches. It is quite possible to get
> sub-second searches for 300,000 items depending uniqueness of term.
> This is the type of thing where a few hours of outside consulting with a
> company that really knows this stuff (like us) will go a long way toward
> solving your problems. This is all database and physics 101, but you
> have to have the knowledge.
this is a shameless promotion, but as Mike Heins started it, I
follow up with my 0.02 cents.
Mike Heins handled (already in 1996) for me 850 000 records (an export of
the complete Baker & Taylor database of books in print) with over 7000
categories (included two levels of subcategorization) on a miserable
Pentium little machine over A MODEM line (!) and set up correct searches,
indexing, static builts and the like in a lower two digit number of hours.
I remember exactly the first time I was able to do the first ISBN
search for a book and IT WAS FAST !! I was too ignorant to recognize at
that time how good that "fastness" really was, but it was so amazing
that I NEVER lost confidence in the whole MiniVend program later on even
after (for completely other reasons) my project failed .
In 1996 MV was still so small that it seemed to be easy to learn for a
complete outsider (like me) to the web design and programmer
community. Today it has grown, got all the stuff you can imagine to
to build equals to the "other famous bookstores" out there and you have
not only Mike Heins alone, but the whole heavy weight of the
Akopia/RedHat team with it.
So, if you are not the poorest of "schmocks" out there (like me) than
my recommendation is to get this help. It's well worth it. Oh, and then
today you could do the whole stuff with XML and that's REALLY