consider an example...

U have emp id from 7001 to 7999.

now as reverse index store values in reverse order...

u will see..id values begning with 1,2,3,4,5...n so..

so now with equaility predicate search for particular id wud be faster than any other indx...
as the range of values with any number(1 or 2 or so) wud be much less than the range of values starting with 7...

okie now with updates goin on in real time systems...since indx values wud be spread ... i/o contention wud be minimal....

Abhaysk you said...
"it wud be useful were in u have a column with values in ascending/descending order....& the values in index wud be spread accross..due vaules stored in reverese struct..."

As far as i know all indexed columns by nature have values that are either ascending or descending, big deal, if they weren't i use a bit-map index ideal for columns of low cardinality! If the values were spread as you say i would have used a B-tree not reversed, reference my last message. Not sure what you are trying to say here, perhaps you could clarify.
this was just to support above 2 statements.

however it has as well disadvantages over b tree normal.

i love Jmodic to throw more light on this.

Abhay.