Graham wrote:
> I don't seriously believe any aggregator that uses the content hash
> approach would survive very long in the market place without being
> buried under user complaints. Most (the one's I know of) either use
> identifiers or failing that some subset of the elements.
The "identifier-based" approach works wonderfully for a single user
aggregator which has a "feed" focus as most personal aggregators do today.
However, as pointed out elsewhere, the identifier-based approaches don't
work in cross-feed duplicate detection because of the ease with which denial
of service attacks can be launched.
If a subset of elements is to be used in duplicate detection, what
is that subset? Can/Should this subset be commonly known? It seems to me
that it is important enough to the atom-ecosphere that it might even make
sense to have it in the spec as an important interoperability note. i.e.
"Entries will be considered duplicates if...."
> Shrook uses an adaptive subset that means that if one element is
> unreliable it uses the others.
Can you describe your algorithm? What do you consider "unreliable"
to mean?
bob wyman
