On Wed, 5 Dec 2007 23:49:05 +0000, Dr R. Sanderson wrote
> What does it mean to have an updated timestamp for a search result?
We have actually time stamps on search results...
When one searches: the result set and its query get stored.
When a request arrives for the "exact same query" it checks the available
cache and if its matches and its date is after when the target was last
modified then it'll return the same result and not do another search.
Its really a simple if-modified-since... In a web environment we can put
a web cache into the chain to even better offload things..
> Does it change every time? What if you just repeat the same search
> and none of the entries have changed? What if you repeat the same
> search, the results are the same, but the sort is different?
In my engine a query structure is defined by:
- the query expression automata-- different sentences in different query
languages (for that matter even paradigms) can produce the same automata.
- the sort
- the score normalization
- a maximum number of elements we'd accept (or unlimited)
These together define a query in my system so change the sort and its
a different query.. Sometimes one can covert from different sorts or
normalizations to another but not always. This depends upon a host of
other system parameters.. In other words don't assume that a result set
sorted by score and normalized by algorithmX can be converted without
access to the database to sorted by date by even the same normalization,
not to mention another (part of the issue is implementation specific to
address things like the sort for two items that have the same X, here
That's the search but the retrieval side is not completely decoupled..
and it can include a request as structure (as complete or partial paths
of ancestors of hits and descendants thereof).
That's my real life system and not a theoretical construct..
Edward C. Zimmermann, Basis Systeme netzwerk, Munich
Office Leo (R&D):
Leopoldstrasse 53-55, D-80802 Munich,
Federal Republic of Germany