Re: [TSP] Socket Based Protocol rationale

[tho <thomas.fossati@xxxxxx>]

Hi Zoltan,

On Fri, Dec 20, 2002 at 11:47:42PM -0000, Zoltan Glozik wrote:
> - partialMsgRep: Can be sent as a response to a tsaMsg or a pollReq. The
> 'next polling reference' and 'time-to-check-back' is straightforward, the
> 'DER encoded TSA message' should contain a TimeStampResponse with a Waiting
> status.
> - finalMsgRep: The same as the previous one except that the
> TimeStampResponse is missing.

Your answer is a reasonable one, thank you, though see my comments below.

> As a side note, I think the introduction of the polling was a good idea. If
> the time source is not available or the server is simply overloaded the
> client can close the connection and poll later.

This is something more black-magic related :-) A busy server or one whose
clock isn't - for some reason - available, hardly can esteem the moment when
things will be ok (i.e. a judicious value for time-to-check-back).

Anyway, my argument here is that the use model for CMP, where a request
can explode in many small pieces each of which can be processed independently
- a partialMsgRep makes sense there - is not valid for TSP, where a request
is basically atomic and should be processed atomically.

IMHO the whole pollRep and partialMsgRep brigade should not be invited
to the party.


Thomas