[Fenfire-dev] pointers article

[Janne Kujala]

Some comments (some edited/commented on by benja) on the 
manuscript: 

Abstract
--------

- what are filesharing systems? (not defined anywhere)

- do they really not allow updating?

- "propose pointer records" *for what*? (should clearly 
  mention what problem they solve)

- why not for dynamically generated pages?
  (it would be nice to have the same benefits for them, too)

Introduction
------------

- oceanstore's primary replica system for *generating* the 
  heartbeats seems more specialized than the proposed pointer 
  systems; perhaps should emphasize the difference

- are there any downsides on not having a primary replica?
  (why did oceanstore opt to have it?)

Pointer records
---------------

- why describe the pointer charter so vaguely --
  why not simply say, "The charter also includes information
  to distinguish the pointer from all other documents
  owned by the same entity, e.g. a random bitstring."

The Storm data model
--------------------

- how and where is the metadata (MIME type) stored?
  (do all p2p systems provide for metadata or
  is that new functionality implemented in storm?)

- "publish mappings" sounds like publishing funtions and
  not single correspondencies

Conclusions
-----------

- Evaluation: dos it work in practice? Is there nothing
  to mention about experiences in practical use?

- Mention p2p anonymity, privacy, bandwidth and other issues, 
  both pros and cons (especially, because p2p web is the main example!)

- How to manage keys, signatures, etc. in p2p web?

- "fake" versions: "a time-stamping mechanism"?
  how does that relate to the timestamp of the pointer record?

Janne