>> 1. A strong convention is needed for the descriptors, so that there is
some
>> chance of finding the content. I'm not sure that this can be left to
chance,
>> and maybe at least the top levels need to be defined - maybe something
like
>> the Dewey Decimal system.

>Yes, a first stab at specifying some descriptor conventions can be found
at:
http://freenet.sourceforge.net/index.php?page=key_const

I saw that, and the comment extending it for classical music. However that
is only a small part of possible content. How would I describe episode 113
of the Simpsons, or a PGP key ring or any of the other things I might want
to access?

>> 2. Some way of finding out what content exists for a specific category,
for
>> example an intelligent search agent.

>There is right now an ongoing argument about how to do searching. =)
I'd be happy to join in. Ive done some work on using fuzzy logic for TV
program guides that might be relevant.

>> 3. Some form of discipline for the namespace. First-come-first-served has
>> been a disaster for the web, and  some mechanism is needed to prevent
abuse,
>> for example by flooding all of a category with placeholder documents.

>There are a couple of competing proposals here, either by using some type
>of user confirmation/voting mechanism to decide which document has the
>"right" to a name, or by a two-level naming system where a
>human-understandable key leads to a set of references to approximately
>unique non-understandable keys (corresponding to content hashes of
>different documents, for example).

There is a corresponding debate in IETF, and an extra level buys you some
things, but still leads to the problem of resolving conflicts and preventing
flooding in the human-understandable space. It's a hard problem.

>> 4. "distance" needs to take into account strength of the connection

>This has been talked about a little, but there is also the problem that you
>don't just want to go to the closest node to look for the data, you want to
>go to the node which is close and is most likely to have the data.
>Balancing these two issues is a problem.

I agree. My point is that you need a QoS parameter as well that defines
things like speed of link and resilience of node- not all links or nodes are
equal. Do the methods used in distributed caches like Squid help any?

>> 5. Can streaming media be supported? If so how?

>I don't see why not, in principle.  When an insert of a streamed file
>reaches a node, it can make a note of it right away in its data store
>(rather than wait until the file is complete).  If someone else requests it
>while the file is still being read in, the request is fed out from the top
>of the file which has already been saved.

>Once the file is complete, of course, no alteration is necessary.  Clients
>will just request it as usual and start interpreting it immediately as the
>file arrives.

I like that. I want to distribute TV pictures. You might want an option of
not storing it at all - think of a Webcam, for example. You get the current
version or none.

Jack

>theo


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to