[topicmaps-comment] RE: [topicmapmail] Subject Indicator Fingerprints[WAS: Genetic PSIs [Re: Topic Map domain, paradigmatic PSIs ...]]

[jean delahousse <jean.delahousse@mondeca.com>]

Hello all,

I obviously like Bernard idea, I'd like to add some comments :

I think the idea is great when the knowledge is huge, fuzzy, decentralized -
a very good example of replicated PSI that converge is the name, author,
style of songs on the P2P server : Kaaza - anyone can give any wrong name to
a song, with mispelling, not indicating the singer or the style, but finally
when you copy a song and have the choice to take the same song with good
metadata and idenfication or loosy ones, you take the good ones and the good
ones replicate more and more on the network and get more chance to be
replicated. Also when you get mistake in the naming or the metadata, very
often you do the correction locally so the song is well indexed wiht other
song of the same singer or style. Those small correction will be replicated.
So as more people do cleaning and as few people take time to insert error in
naming and metadata, the system create organisation and is able to manage
mistake.

In an other field like organisation of electrical appliances and their
technical documentation for a large company, I don't think this process of
fuzzy replicated PSI will do any good. An extremely controled and quite
centralized catalog of PSI is needed to be shared by the various content
management systems.

Sincerely

Jean Delahousse
Mondeca
CEO
3, Cité Nollez. 75018 Paris
Phone : 33 1 44 92 35 03
Mobile : 33 6 16 90 73 95
Fax : +33 1 44 92 02 59

www.mondeca.com

> -----Message d'origine-----
> De : topicmapmail-admin@infoloom.com
> [mailto:topicmapmail-admin@infoloom.com]De la part de Kal Ahmed
> Envoyé : lundi 8 avril 2002 23:05
> À : Steven R. Newcomb; Bernard Vatant
> Cc : topicmaps-comment; tm-pubsubj-comment; topicmapmail@infoloom.com
> Objet : [topicmapmail] Subject Indicator Fingerprints [WAS: Genetic PSIs
> [Re: Topic Map domain, paradigmatic PSIs ...]]
>
>
> Firstly, thanks to Bernard for bringing up yet another intersting
> "thought
> for the day"!
>
> I can't resist putting on my developer's hat when I read this and try and
> figure out how a PSI could be successfully "replicated" and yet still
> remain identifiable with minimal processing.
>
> I see an issue here between the forms of identification and resolution
> mechanisms so far proposed for PSIs and the pursuit of stability through
> replication, but fortunately, this is not an issue which has not already
> been addressed by those developing decentralized information sharing
> systems and although no expert in this field, I suspect that an
> answer may
> be lurking in some clever application of cryptographic hashes to a
> published subject indicator. Some application such as this would allow a
> subject indicating resource to be copied wholesale from the original
> publisher to another site, (thus reassigning subject indicating
> resource's
> address), yet its origin could still be securely and consistently
> identified (at least in theory...as I say, I'm no expert on this subject
> ;-). An application could then determine a PSI's fingerprint by
> retrieving,
> validating and finally digestifying the content - assuming that
> all is OK,
> this would lead to a match. This might seem like a lot of
> overhead, but in
> a distributed system, it is the price of information security -
> of course,
> in a more controlled environment, various short-cuts could be
> applied. For
> example, a "subject indicator directory service" might retrieve and
> validate the multiple addresses for the same subject indicating resource
> and then publish a topic map which consists merely of alternate subject
> indicating resource addresses for the subjects it aggregates.
>
> While the forgoing is somewhat fuzzy, my feeling is that the PSI group
> should consider that a location-dependent subject indicating resource
> address (such as an HTTP address) is not necessarily a barrier to the
> robust replication of subject indicating resources, but that in order for
> it not to be a barrier, another level of technology may be required for
> such a location-independent scheme to be trust-worthy in an open system.
>
> Cheers,
>
> Kal
>
> _______________________________________________
> topicmapmail mailing list
> topicmapmail@infoloom.com
> http://www.infoloom.com/mailman/listinfo/topicmapmail
>