RE: Time-stamp server. TimePrecision info

[Michael Zolotarev <mzolotarev@xxxxxxxxxxxxxxxx>]

Michael,
Technically you are right, and I certainly agree that monotonicity is 
absolutely essential, and assumed. It is accuracy and precision which I am 
trying to address here.

I believe that when considering what might go into a time stamp, we should 
think in terms of what a user would NEED to know. After all, what matters 
for a client is
1) how trustworthy is the TSA in general, and
2) how accurate is the time in that particular time stamp.

Presenting a customer with extra details, like splitting the quality of the 
time into its precision and accuracy, may unnecessarily overcomplicate the 
issue, without giving much information anyway.

A reference to the source(s) of time data, available from the TSA's policy, 
should be a starting point for anybody who wants to know more about 
characteristics of the time.

Thinking of monotonicity:
Should the TSTInfo::serialNumber field be OPTIONAL? In addition to serving 
as just an extra proof of monotonicity, which is assumed anyway, the 
serialNumber may be the only way to resolve a 'who was the first' argument. 
If two requests arrived virtually 'simultaneously', the TSS will generate 
two time stamps with the same time value. The probability of it depends on 
the resolution of the time source used. It may be critical for the clients 
to know who was actually the first (thinking of on-line betting, auctions, 
etc). This turns the serialNumber into a mandatory field, doesn't it?

Michael Zolotarev
Technical Architect, Baltimore Technologies Ltd (Australia)

-----Original Message-----
From:	Michael Sierchio [SMTP:kudzu@dnai.com]
Sent:	Monday, March 29, 1999 3:32 PM
To:	mzolotarev@baltimore.com.au
Cc:	'ietf-pkix@imc.org'; Simon Laing; Andrew Shellshear
Subject:	Re: Time-stamp server. TimePrecision info

Michael Zolotarev wrote:
> ...
> for example, the policy might state: "The time returned will be accurate 
to
> within 100ms, and if the time-stamp server cannot
> provide this level of accuracy, it will not issue a time-stamp."

It seems to me that there are three qualities of interest for
timestamps, in increasing order of importance -- precision, accuracy,
and monotonicity.  The last is essential,  and perhaps more important
than the other two.