[Gzz] Re: [ba-unrev-talk] Starting Point for Collab Tool

[Eric Armstrong]

All in all, this is really good news. I understand that refactoring
puts things in a bit of flux at the moment, but the result of it all
will be an excellent segmentation that lets Storm get used in a
variety of projects. It's really rather thrilling.

On the mathematical side, I woke up the other day and realized
that if I want to track who ever wrote anything, I'll need an identifier
big enough to identify anyone who has ever lived (for example,
Shakespeare), and to identfiy who makes changes, I'll need one
big enough to identify anyone who ever *will* live. Now, that
identifier is starting to get pretty huge.

Then, I'll need an identifier for every paragraph in every document
on every computer that ever was or will be constructed. These
are really big numbers now. And when someone corrects a typo
on some page somewhere, that paragraph will need to be versioned,
and the huge-long-identifier of the perpetrator will need to be stored,
along with a date, just to track that one-character change.

I began to think that, mathematically, perhaps true collaboration in
the form of "global knowledge sharing" was impossible, and that
the best we could do would be some sort of "local cluster", within
which true sharing could go on.

I look forward to the results of any thinking you guys have done on
this subject, or any research you've managed to find that deals with
it.

Alatalo Toni wrote:

> On Fri, 21 Feb 2003, Eric Armstrong wrote:
>
> > Your article has me really intrigued. I've only mangaged to
> > leaf through it so far, but I've seen enough to know that
> > targets the major, critical impediment.
>
> good to hear that it managed to communicate - i hope you don't mind cc'ing
> this reply to the actual developers .. this way you'll get better answers
> to the questions.
>
> > The next step is to evaluate how well it does so. Have you
> > given any thought the mathematical difficulties of the process?
> > (Identifiers have to get simply huge!)
>
> one thing about the evaluation: feel free to get a copy from
> http://savannah.nongnu.org/cvs/?group=gzz and see it work!
> there are some bleeding edge dependencies especially on the GUI side,
> but much less for the storage module.
>
> the project has been using Storm for about 1,5years now, as mentioned in
> the article. the p2p networking is under development, but the basics, like
> creating the block identifiers (and verifying them) are there and work.
>
> i personally don't know much about it on the mathematical level, so the
> other authors should correct me here, but my understanding is that SHA-1
> hashes are not awfully difficult to deal with. i need to look more into
> it, though, also to analyse the applicability in small devices. related to
> p2p networking there are of course several issues, that some of the
> authors are researching.
>
> evaluation is definitely one aspect to develope in the article, too, so i
> think it's good to have these comments.
>
> > I know Eugene and some others have given this matter some
> > thought. Have you sent a copy to Eugene, or to the Yak group?
>
> no, not yet - i guess it would be good to hear reviewer's comments first.
>
> also, this is bad timing for the reason that the project is in quite a
> flux right now: we agreed on a set of new names on Friday, the whole is
> now split to several modules and there are rewrites going on due to the
> change of the structure (from zz to rdf, but that's not related to Storm).
>
> a definite good thing in that reorg is that it will be clearer to use the
> modules from outside of the project. so e.g. Storm is a whole of it's own
> and can be used independently of the other products of the ex-Gzz project
> (Fenfire is the new name, hence it's the Fenfire Storm now). so in a week
> or two we should be on firmer ground -- just wanted to reply to you
> immediately as a comment to the promising post on the unrev-list.
>
> looking forward to what becomes of this,
> ~Toni