p2p summary (kind of)

Joe Armstrong joe@REDACTED
Thu Feb 14 13:55:55 CET 2002


On Thursday 14 February 2002 1:00 pm, Vlad Dumitrescu wrote:
> Hi,
>
> I'd like to check if my understanding of terms is in accordance with yours
> or not.
>
> In my world, an UUID is a reference to something, like a C pointer.
>

I was thinking of the meaning that the JXTR spec. uses.

Quote:

  JXTR uses UUID, a 128 bit-datum to refer to a peer ... omitted ..
  It become significant only after it is bound to other information ...
  We assume and expect that sophisticated naming and binding services will be 
  developed for the JXTR platform

EndQuote.

Which I take to mean it's a 128 bit thing that points to something but 
*finding the thing it points to is not defined* - but this to me is the knux 
of the problem.

 
> A Dewey category is a description of a group of entities, meta-data; like a
> type declaration for a C entity.
>

It's a standard for names.

> Therefore, they are two completely different beasts, each with its
> usefulness. I think they are both needed (not necessarily in the form of
> UUIDs and Dewey numbers, but as a reference to something and a type
> description of that something). It's hard to imagine how things would work
> without any of them.
>

Yes you need both.

> In the p2p world, UUIDs (or other kind of unique references) can identify
> specific users, or specific services, or specific global data. Dewey
> numbers (or other kind of meta-data) can be used in searches, as a
> complement to regexps :-), and in describing the thingie that an UUID
> refers to.

My problem (with JXTR) is that if you read the standard for ideas whenever 
things start getting interesting (like how to find the peer where a UUID came 
from) the text get vague and says (this is out of scope, or, an implementation
issue, or, they "expect that something will be developed which will ...")

It's a kind of empty framwork with a terminology - admittedly it clarifies
various design issues (like breaking things up into different prototcols,
the discovery protocol, the peer resolver protocol etc.) but as you dig into 
the detail you find there is no *helpful* detail in the sence that the detail 
provides insights  into how to do things.

I guess this is getting very off-topic (sorry) - we'll move to the wiki


> best regards,
> Vlad
>
> _________________________________________________________________
> MSN Photos är det enklaste sättet att dela ut och skriva ut foton:
> http://photos.msn.se/Support/WorldWide.aspx



More information about the erlang-questions mailing list