RE: [dss] Timestamping

[Trevor Perrin <trevp@trevp.net>]

At 06:45 PM 3/21/2003 -0500, you wrote:

>Trevor,
>
>
>The only protocols defined in the ISO documents are the
>timestamp request protocol and the timestamp verification protocol.
>No additional protocols specific to linked timestamp tokens
>were defined.

[..]


>Aggregation of timestamp token requests may be performed according
>to the TSA policy, for example every 500ms or after at most 512
>new timestamp requests have been received, whichever comes first.
>According to 18014-3, aggregation is viewed as a way for the TSA
>to scale up its throughput, so no additional protocols were
>defined for "come-pick-it-up-in-an-hour" type of messages.
>Aggregation is not necessarily the same as a timestamping round
>the way the latter term is sometimes used in the literature.
>18014-3 makes no assertion about the ordering of the timestamp
>requests processed in a single aggregation operation
>(i.e., the TSA could have reordered them according to his fancy),
>while some of the papers describing algorithms using timestamping
>rounds make assertions about the ordering of the requests
>within a given round.
>
>Freely paraphrasing from 18014-3:
>All timestamp tokens generated by an aggregation operation
>shall contain the same time value.  The TSA shall maintain
>a repository of the linked values resulting from each aggregation
>operation.  This repository is used for the verification of the timestamp
>tokens issued by the TSA.  The integrity of the repository
>is ensured by requiring that the TSA publishes in a timely fashion
>values derived from the repository of linked values computed above;
>different publishing schemes are recommended.  All communications
>between a requestor/verifier and such a TSA shall be over a data
>integrity and origin authentication protected channel.
>
>Dimitri


I thought one goal of aggreggation was that the TSA would publish the 
values of each aggregation round, so they could be widely archived by 3rd 
parties, and thus the TSA would not have to be absolutely trusted.  I 
assumed this would mean an aggregation round once a day or something.

If the server is doing an aggregation round every 500ms, thats 170,000 a 
day, or 63 million a year, which is I guess too many to widely publish.  So 
it appears from the last paragraph above, that the TSA publishes in a 
timely fashion "values derived from the repository of linked values 
computed".  So how does that work?  Is that like a super-aggregation round 
on the sub-aggregation's round's values or something?

In any case, how does the TSA publish these values, and how does someone 
verifying the timestamp know where to retrieve them?  Does that require 
protocol and and data format support?

I'm still just wondering if this is too complicated for us to spec out 
fully enough that different implementations will interoperate (i.e. that 
any timestamp-verifying service will interoperate with timestamps produced 
by any TSA).

Trevor