[whatwg] Parsing RFC3339 constructs

[Julian Reschke]

Ian Hickson wrote:
> On Fri, 5 Jun 2009, Julian Reschke wrote:
>> Ian Hickson wrote:
>>>> Michael(tm) Smith wrote:
>>>>> It seems pretty clear that there isn't anything else to refer to for 
>>>>> the date/time parsing rules -- but to me at least, specifying those 
>>>>> rules seems orthogonal to specifying the date/time syntax, and I 
>>>>> would think the syntax could just be defined by making reference to 
>>>>> the productions[1] in RFC 3339 (instead of completely redefining 
>>>>> them), while stating any exceptions.
>>>>>
>>>>> [1] http://tools.ietf.org/html/rfc3339#section-5.6
>>>>>
>>>>> I think the exceptions might just amount to:
>>>>>
>>>>>   - the literal letters T and Z must be uppercase
>>>> Any technical reason why they have to?
>>> Not really. We just need a separator.
>> So why make it different from RFC 3339?
> 
> Limiting the syntax to the simplest possible syntax was an intentional 
> design choice intended to ease the burden on implementors and authors. In 
> practice, pretty much every time we've made syntax case-insensitive, we've 
> ended up having trouble because of it.

If this was a totally new syntax, I would agree.

But as something based on ISO8601 (and thereby also RFC 3339) it appears 
to be a bad idea to make it less compatible just for that reason.

>>>>>   - a year must be four or more digits, and must be greater that zero
>>>> "a year must be four or more digits" -- sounds like an alternative 
>>>> format that an additional RFC, updating RFC 3339 could specify.
>>>>
>>>> "must be greater that zero" -- that's not syntax :-)
>>>>
>>>> So yes, I think referring to RFC 3339, even if it's just a narrative 
>>>> mention, would be good.
>>> Why?
>> Because it explains to readers how this is different. That is important
>> because it's natural to look for existing libraries to parse date formats.
> 
> The HTML5 spec defines exactly how to parse dates. Implementors are 
> required to implement what the spec describes, so reusing libraries is 
> implicitly not likely to be useful here. RFC3339 isn't even a particularly 
> important one in the grand scheme of things (ISO8601 comes to mind as a 
> much higher-profile example).

I think it's unfortunate that HTML5 doesn't allow using an off-the-shelf 
parser. But if it doesn't, and the temptation *will* be there to use 
them, I'd recommend stating it very clearly.

> I'm certainly not proposing to go through every date format spec and 
> explain how the rules in HTML5 differ from those rules. That is the kind 
> of material that belongs in support documents.

BR, Julian