- From: Francois Daoust <fd@w3.org>
- Date: Fri, 06 Jun 2008 14:31:57 +0200
- To: Jo Rabin <jrabin@mtld.mobi>
- CC: MWI BPWG Public <public-bpwg@w3.org>
Indeed, in all cases, the important point is that we state that the
User-Agent string may be extended.
I keep thinking about the Accept header though and the possibility that
typical libraries include a space after the comma. This is rather
editorial stuff though, so I propose the following:
* Let's go to LC with the change on the User-Agent as you decide it as
soon as possible
* While in LC, let's see if that needs to be refined in a
consistent-and-in-the-spirit-of-HTTP way and, at the very worst, I'll
make a LC comment with a proposed and polished rewording. It would be a
minimal editorial change anyway with no impact on further progress of
the doc along the Rec track.
Francois.
Jo Rabin wrote:
> tks Francois,
>
> I think that this is more like naming the angels on the head of the pin,
> rather than merely counting them (one of Rotan's) - my editorial
> discretion says, for the reasons cited, to go for the formula that says:
>
> "starts exactly as follows (and may be extended in accordance with
> [HTTP] [Section 14.43, User Agent Header]"
>
> but will leave it to COB London for others to express their views before
> issuing a revision.
>
> Jo
>
> On 06/06/2008 11:47, Francois Daoust wrote:
>> +1 (like in crazy in love) for the "starts exactly as follows (and may
>> be extended in accordance with [HTTP] [Section 14.43, User Agent
>> Header]", especially if we're to be consistent with the Accept and
>> Accept-Charset headers and would have to change the wording to an
>> in-the-spirit-of-HTTP one for these two headers as well.
>>
>> Matching strings exactly and matching beginning of strings is easy.
>> Having to apply BNF grammar parsing is less trivial. I can see that a
>> mobileOK checker implementation might use some HTTP library to send
>> requests that doesn't quite allow to control the headers to match
>> these strings entirely. I don't think that's really going to ever
>> happen in practice though.
>>
>> Again, I'm not willing to block/postpone anything. It's a +1 to the
>> text that follows the Occam's Razor principle IMO. But it's not a -1
>> to the alternative text. If no one else has any strong view one way or
>> the other, I leave the final decision in the hands of the editor with
>> pleasure.
>>
>> Francois.
>>
>>
>> Jo Rabin wrote:
>>> I'd like to draw this to a conclusion to I can produce a new draft
>>> today ...
>>>
>>> I think we need to decide whether we mean send a User Agent header
>>> which starts exactly as follows (and may be extended in accordance
>>> with in accordance with [HTTP] [Section 14.43, User Agent Header]) "
>>> or whether, more in the spirit of HTTP, we say that when parsed in
>>> accordance with the BNF defining the User-Agent Header, the value is
>>> a <product> set to the value described followed by a <comment> set to
>>> the value described, followed by anything that conforms to the
>>> structure of the User-Agent header.
>>>
>>> Note that we also have
>>>
>>> #
>>>
>>> Include an Accept header indicating that Internet media types
>>> understood by the default delivery context are accepted by sending
>>> exactly this header:
>>>
>>> Accept:
>>> application/xhtml+xml,text/html;q=0.1,application/vnd.wap.xhtml+xml;q=0.1,text/css,image/jpeg,image/gif
>>>
>>>
>>> #
>>>
>>> Include an Accept-Charset header indicating that only UTF-8 is
>>> accepted by sending exactly this header:
>>>
>>> Accept-Charset: UTF-8
>>>
>>> #
>>>
>>>
>>> so whatever we decide perhaps we should be consistent. In the case of
>>> Accept and Accept-Charset - these are very definitely not extensible,
>>> though perhaps we should allow for trivial syntax variations. I don't
>>> feel strongly about it.
>>>
>>> I think it is clear that we _don't_ mean:
>>>
>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>> http://www.w3.org/2006/07/mobileok-ddc (I am a malicious crawler))
>>>
>>> We _do_ mean
>>>
>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>> http://www.w3.org/2006/07/mobileok-ddc) (I am a malicious crawler)
>>>
>>> Jo
>>>
>>> Not wishing to bore with the detail but we did say EXACT:
>>>
>>> (To my mind is is a little bit unclear exactly where LWS is allowed,
>>> it would appear to my untutored reading that any token, separator or
>>> quoted string forming part of a field value may be preceded by (or
>>> actually followed by) LWS.)
>>>
>>> message-header = field-name ":" [ field-value ]
>>> field-name = token
>>> field-value = *( field-content | LWS )
>>> field-content = <the OCTETs making up the field-value
>>> and consisting of either *TEXT or combinations
>>> of token, separators, and quoted-string>
>>>
>>> User-Agent = "User-Agent" ":" 1*( product | comment )
>>> product = token ["/" product-version]
>>> product-version = token
>>> token = 1*<any CHAR except CTLs or separators>
>>> separators = "(" | ")" | "<" | ">" | "@" | "," | ";" | ":" | "\" |
>>> <"> | "/" | "[" | "]" | "?" | "=" | "{" | "}" | SP | HT
>>> comment = "(" *( ctext | quoted-pair | comment ) ")"
>>> ctext = <any TEXT excluding "(" and ")">
>>> quoted-pair = "\" CHAR
>>> LWS = [CRLF] 1*( SP | HT )
>>> CRLF = CR LF
>>> CR = <US-ASCII CR, carriage return (13)>
>>> LF = <US-ASCII LF, linefeed (10)>
>>> SP = <US-ASCII SP, space (32)>
>>> HT = <US-ASCII HT, horizontal-tab (9)>
>>>
>>> TEXT = <any OCTET except CTLs, but including LWS>
>>> OCTET = <any 8-bit sequence of data>
>>> CHAR = <any US-ASCII character (octets 0 - 127)>
>>> CTL = <any US-ASCII control character (octets 0 - 31) and DEL (127)>
>>>
>>>
>>> On 05/06/2008 21:09, Francois Daoust wrote:
>>>> Jo Rabin wrote:
>>>>> My understanding of the discussion was that we started from that
>>>>> and then went on to what is discussed below - which I believe is
>>>>> what is said in the resolution. I prefer what we have now, as it
>>>>> seems, do you.
>>>>
>>>> Well, I prefer as well.
>>>>
>>>>
>>>>>
>>>>> I think it better not to specify the "string" it starts with
>>>>> because that in fact implies something about the spaces and other
>>>>> theoretically insignificant things to my mind.
>>>>
>>>> OK, it's just that I still don't see why we can't impose the
>>>> beginning of the string.
>>>>
>>>> Knowing that:
>>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>>> http://www.w3.org/2006/07/mobileok-ddc)
>>>> (5 spaces between the product token and the comment)
>>>> ... may be a valid User-Agent string (it may not, I haven't checked)
>>>> should not prevent us from saying that, for us, it must be:
>>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>>> http://www.w3.org/2006/07/mobileok-ddc)
>>>> ... and then implementers may add whatever they want to the end of
>>>> the string.
>>>>
>>>> That being said, since the first product token is perfectly defined
>>>> as "W3C-mobileOK/DDC-1.0", the proposed text is totally fine!
>>>>
>>>> Francois.
>>>>
>>>>>
>>>>> Jo
>>>>>
>>>>> On 05/06/2008 17:11, Francois Daoust wrote:
>>>>>> Jo Rabin wrote:
>>>>>> [...]
>>>>>>> Proposed Text:
>>>>>>>
>>>>>>> Include a User-Agent header indicating the Default Delivery
>>>>>>> Context by sending a product token set to "W3C-mobileOK/DDC-1.0"
>>>>>>> followed by a comment set to "(see
>>>>>>> http://www.w3.org/2006/07/mobileok-ddc)". These may be followed
>>>>>>> by any number of other product tokens or comments in accordance
>>>>>>> with [HTTP] [Section 14.43, User Agent Header]. The minimal User
>>>>>>> Agent header is:
>>>>>>>
>>>>>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>>>>>> http://www.w3.org/2006/07/mobileok-ddc)
>>>>>>
>>>>>> OK, I don't mean to be picky on this, but I probably lost myself
>>>>>> in the BNG dicussion. My point is that I thought we agreed that
>>>>>> the following was valid:
>>>>>>
>>>>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>>>>> http://www.w3.org/2006/07/mobileok-ddc; my comment)
>>>>>>
>>>>>> -> Based on the proposed text, it's not. Actually, I don't mind
>>>>>> either way, with a slight preference for it to be invalid anyway,
>>>>>> but I just want to make sure this is what was discussed and agreed.
>>>>>>
>>>>>> It's good if it's invalid, although I don't quite see in that case
>>>>>> why we don't simply state:
>>>>>> "Include a User-Agent header indicating the Default Delivery
>>>>>> Context by sending a header that starts with:
>>>>>> User-Agent: W3C-mobileOK/DDC-1.0 (see
>>>>>> http://www.w3.org/2006/07/mobileok-ddc)"
>>>>>> But that's probably not rec-friendly enough...
>>>>>
>>>>
>>>
>
Received on Friday, 6 June 2008 12:32:36 UTC