My favorite feature was 'near' queries. One could project n-dimensional spaces to any arbitrary lower dimension with the right semantic and spatial distances.
With Altavista, the limiting factor was the searcher. And my nostalgia that over-estimates the size of their search database.
> One could project n-dimensional spaces to any arbitrary lower dimension with the right semantic and spatial distances.
I'm incredulous. Doing this well for unstructured text would be worth a lot in many industries today, let alone the entire web. If they did this well, they'd still be alive and well in another form.
Maybe a smaller search corpus made it feasible, but this is an accurate description of AV's functionality. From http://jkorpela.fi/altavista/ , which is worth reading in its entirety:
The operator NEAR ensures that both are within ten words of each other in the resulting documents. The operator NEAR binds less tightly than the operator NOT and associates to the left.
With Altavista, the limiting factor was the searcher. And my nostalgia that over-estimates the size of their search database.