Re: [Bug-gnupedia] Re: A Detailed Proposal - Mk I

[Imran Ghory]

On 20 Jan 2001, at 9:33, Mike Warren wrote:

> "Imran Ghory" <address@hidden> writes:
> 
> > I think that unique IDing should be done on the content of the
> > article not on data such as the server it was submitted to. After
> > all we want two articles which are the same to be fingerprinted the
> > same regardless of external facts such as which server it has been
> > submitted to.
> > 
> > I think MD5 (rfc1321) fingerprinting of the article would be the
> > best way to do this.
> 
> Even simpler, we could just sequentially assign 128 (or 256 or 512...)
> bit integers to the articles as they arrive at the ``central'' server,
> especially since the articles' content will change over time...

Do we really want a central point of failure ?

The first version should kept, fingerprinting the original would 
encourage servers to keep older articles.

Alternately as I suggest before updates to an article could be 
stored as diffs. If the diffs get too large it would be a good sign that 
the article is hardly the same as the original and should be given a 
new fingerprint to indicate such.

Imran Ghory