Re: ABNF for HTTP headers

2010/5/20 RaphaŽl Troncy <>:
> Hi Silvia, all,
> [That has again been discussed today on IRC, so I'm answering to this mainly
> for the record]
>> for media fragment URIs, the spec defines temporal specs as follows:
>> t=npt:10,20
>> but in the HTTP header examples, we have: t:npt=10,20 .
> Indeed, the syntax is different, but should not be seen as curious. Our
> rationale is to follow respectively the URI fragment syntax and the HTTP
> header syntax (e.g. a normal byte ranges request). There is no reason that
> the URI syntax should be the same than the header syntax. And there will be
> no code optimization anyway, since what can be written in a URI and what can
> be written in a header have different constraints (think about the %-encoded
> strings). Since there must be some re-writing that should happen anyway, we
> can make the header syntax very different that the URI syntax, as soon as we
> have a good reason to do so. Our 'good' reason is to align with the current
> syntax of the header :-)

I've defined the ABNF in this way. I just stumble across this every
single time I look at the headers and the URLs they encode.


Received on Thursday, 20 May 2010 01:03:53 UTC