[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]

ent developments



I was trying to figure out how to test the bert canopy this morning
when I realized that it is completely independent from the ent.  Since
I add information to the canopy in a pseudo-balanced fashion, I don't
need to rebalance, so I don't need backpointers into the ent.  I
implemented the balanced insertion last night, so all that remains is
adding the update routine to the ent (tonight).  Then sensors....

Remember the intra-ent blue planes?  This is the part of the canopy
(undeveloped) which could represent endorsements on sections of
documents.  I thought of a way to represent that using only a little
bit of extra bookkeeping and the current canopy.  As an interesting
section of a document, I'll have a link endorsed by its linkType.
The link is essentially treated like a separate document from the
canopies perspective.  Therefore, the subtree southward of the link
combines the endorsements on the document and on the link.  Once the
backfollow operation gets to the link (but not all the way to the top
of the document), it discovers that the link itself has endorsements
and adds that to the set of discovered things.  I just returned from
MarkM's house where we discussed this in simlar detail to try to
discover holes.  We deliberately avoided designing the mechanism for
'realizing' that the link itself had endorsements (or might go
implement this), but I've got some real good ideas....  If we discover
that backfollowing filtering by endorsements on documents and on
linkTypes is too slow (or similar operations), we can add support
directly now (and we will certainly do so eventually).  

Finally, we determined that canopy for sensors would be even easier
than the bert canopy due to the assymetry of creation in the ent.
Finally, we decided to call crums in the bert canopy bertCrums and
crums in the sensor canopy sensor crums.  Note that due to a
simplification in the algorithms, brown planes no longer exist (canopy
indexing planes).  Some future version might have to add them for more
sophisticated canopy management, but I expect not.

The ent is gradually being informed....

That's all from the Ent Woods.
dean