[DALI] timestamp format

[Patrick Dowler]

Comments from the TCG review have pointed out either an inconsistency or 
at least poor referencing in the PR-DALI-1.0 document.

** The problem

The DALI document formalises the current restricted usage of ISO8601 
timestamp strings within DAL services. Specifically, this common usage 
(I think first written down in VOResource and coming from FITS) 
specifies that dates are

YYYY-MM-DD

and that date+time are written as

YYYY-MM-DDThh:mm:ss[.sss]

where the T is the literal separator between date and time. This would 
be called a "profile" of ISO8601 since it specifies a subset of the 
standard. However, IVOA practice, which DALI kept, is to always require 
such timestamps to be UTC (detailed language is slightly more specific), 
while ISO8601 explicitly states that the above is interpretted as "local 
time" and a trailing Z is required if it is UTC.

So, common IVOA usage (and as I understand the "ISO8601 profile" in the 
FITS community) is not consistent with ISO8601.

** What to do?

Well, we cannot just say this is a valid "ISO8601 profile" since stock 
software that groks ISO8601 will *not* interpret our timestamp strings 
correctly.

1. Fix the inconsistency with ISO8601 by requiring the Z for UTC.

2. Stick with the current style and

--2a. reference FITS usage instead of ISO8601 and *do not* state this is 
an ISO8601 profile, just a similar syntax

--2b. claim that in the IVOA "local" means the same as UTC (admittedly 
this is kind of sketchy)

--2c. is there a better justification than backwards compatibilty?


If someone can provide the correct FITS reference, that would help with 2a.

Here is a good overview of ISO8601 without having to buy it:

http://www.cl.cam.ac.uk/~mgk25/iso-time.html

-- 

Patrick Dowler
Canadian Astronomy Data Centre
National Research Council Canada
5071 West Saanich Road
Victoria, BC V9A 2L9

250-363-0044 (office) 250-363-0045 (fax)