Ann,

> Leyne, Sean wrote:
> >
> > I don't see the value of this type of index, expect to handle
> > uniqueness.
> >
> Yes - if you were setting up a list with messages, identifying
> messages that are exactly identical lets you detect one kind
> of spam.

I understand, but I see that problem as being a .0000001% percent
problem -- for our users -- at the moment we need to focus on more
important items.

> For another, the hash index is very dense compared
> to a multi-thousand byte string. There are algorithms for
> compressing the trailing characters, but we don't use them, so
> a 4,000 byte key that differs in the 3,000th byte will create
> an entry thats 1,000 bytes long.

Disk space is very cheap!!!


I am reluctant to consider creating a new type of index which has a
single purpose (uniqueness), that can't be used for other purposes.


Sean