Re: Implicit *LWS

On 7 Apr 2008, at 16:41, Henrik Nordstrom wrote:

> mån 2008-04-07 klockan 13:47 +0100 skrev Geoffrey Sneddon:
>
>> Over the past few days I've been looking over RFC2616, and I've found
>> implicit *LWS to be very troublesome. Is this planned to be got rid  
>> of
>> when moving to ABNF?
>
> Yes, it's one of the changes on the ABNF table..

That's what I was hoping. :)

>> - It is defined as being between words and words, or words and
>> separators, but it never defines what a word or a separator is. Is
>> token a word? Is "foo" a word? Is "/" a separator?
>
> The same paragraph defines word as "token or quoted-string", kind of..

That's not clear whether that is a formal definition or an informative  
example of what counts as words. In the real world, it seems to be  
more than that. It appears whitespace is allowed within HTTP-Version,  
for example.

> separators is defined, but does not really match actual use..

So it is just meant to be the separators rule? Ah.

>> - There are all kinds of places where I think it is allowed (pending
>> the above) where it really shouldn't be. As far as I can see, CRLF =
>> CR *LWS LF (where the *LWS is implicit).
>
> How? Neither CR or LF is a word/token/quoted-string...

Well, that is only worthwhile discussing once we know what a word is,  
and under what you're pointing at, it isn't.


--
Geoffrey Sneddon
<http://gsnedders.com/>

Received on Monday, 7 April 2008 18:09:36 UTC