More attributes for indexing?

Hi Steven,

> Steven Bakker writes :
> 
> At the last RIPE meeting there was some discussion about which attributes
> would be useful to keep indices on.  Well, since our office move last
> weekend, I came up with a (rather predictable) one:
> 
>     $ whois -i ad "Lockton House"
> 
>     % ignored attribute "address" since it is not indexed as a reference
> 
>     % no attributes specified for -i query
> 
> 
>     $ _
> 
> It would be nice to have an index on address, although such an index might
> become veeerrrry large, due to the fact that you can specify multiple
> "address" attributes per object; dunno.
> 
> Any thoughts?

>From the technical point of view:

This is very well possible. Indexing the person objects takes about 1.5
hour. This extra indexing might cause 3 times more indexed values (for
person objects only) and thus will probably take about 4.5 hours at this
point in time (yes, indexing times are *not* scaling exponentially
anymore since the last release ;-)). This means that it could be done on
the current machine for the next year and even for a longer term if
indexing is done in the weekend. A on-the-fly reindexing feature could
solve this problem completely. Note that much of this is harddisk
speed bounded and is thus more difficult to speed up then just buying a
faster CPU. One warning though, the perl/dbm package might give problems
earlier in time, which could make it necessary to withdraw this
functionality later in the future.

David K.
---