> Thing 1: InsertRequest messages need an InsertDataLength field.  Nodes can be 
> configured
> to reject InsertRequest messages if the size of the data is too big.  This 
> doesn't help
> with DataRequest messages, but it's a start.

I agree, we do need something like this - but as always, it must be
user-configurable.

> Thing 2: The data store needs a maximumItemSize parameter; a data store will
> not even bother trying to store a document that exceeds this.

Agreed again.

> Thing 3: We need to come to an agreement on the partial message format so
> that I can implement it.  I've been sketching out a proposal -- here's the 
> current
> draft:
> 
> Properties for the "message/partial" MIME type
>         I have attempted to follow RFC2046
>         regarding this format as much as
>         possible.
> 
> partial.number
>         Identifies which part this is.  Required.
> 
> partial.total
>         The total number of parts.  Required.
> 
> partial.part<n>.key
>         The (hashed) key identifying part <n>.
>         Required in the first part, suggested in all
>         parts.

We should probably ensure this mechanism can be generalized to include
redirection (like a symbolic link or ShortCut in Windows speak), and
possibly different methods for splitting the data.  The mechanism you
describe should work, but we should ensure that it is as elegant and
flexible as it can be.

Ian.

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

Reply via email to