[Help] Last update at http://inet.nttam.com : Mon Aug 7 22:08:51 1995
Comments for"Maintaining Link Consistency in Distributed Hyperwebs"
Share your thoughts on this paper with the Internet.
Previous Comments and Replies
- 05/23/95: from
I really enjoyed reading your paper. I think this
algorithm would have many uses in addition to its usefulness
in maintaining link consistency. I had a question which led
me to a proposed addition to the p-flood algorithm, and I
hope that you will take the time to enlighten me.
Why do you choose the ordering of reverse-domain name for
your circle of primary server-links instead of a more
generic and possibly more efficient arrangement such as
would be produced by a greedy algorithm? That is: to
construct your server-circle at first, you ask server
number 1 who his favorite neighbor is (which other server
he thinks he has the best connectivity with and you hope he
has good reasons for his decisions. He might value uptime
in his neighbors as well as bandwidth.) Let's say he picks
server number 2. Then you ask server number 2 who his
favorite neighbor is who is not server number 1, and he
picks server number 3. You ask server number 3 who his
favorite neighbor is who is not server number 1 or server
number 2, and so on.
When adding new servers later, you can allow the new
server to pick his two favorite neighbors which are
currently connected and then squeeze in between them.
(Sorry for the gross anthropomorphism, but I think it
communicates my idea...)
It also seems like the algorithm could be improved by
having a second circle, which could be generated by asking
each server to pick his 8th-favorite neighbor who is not
already on the second circle. When p=2.0, only those two
circles are followed, and when p> 2.0 then the random shots
come into play.
(I base this idea on intuition, and on the principle that
a single level of redundancy is a whole lot better than none
at all. Perhaps you have done quantitative research which
would prove or disprove this idea.)
Thanks for your interesting paper!
- 05/04/95: from
Dr. Leon James
I wonder if your approach can solve a maintenance problem for
virtual superdocuments in generational cybercommunities.
Members interlink each other's reports with two-way links,
so it makes a web that can be used as an index of social
communicative acts in cyberspace. Please take a look at my
paper at this conference (it's no.205, also listed under
authors) and let me know if your algorithmic approach (which
I confess I do not understand) might be applied to this type
of situation -- which I predict will become an important
activity in the future.