> On 11 Sep 2018, at 13:53, Sven Van Caekenberghe <s...@stfx.eu> wrote:
>
> Hi Paul,
>
>> On 11 Sep 2018, at 06:02, PAUL DEBRUICKER <pdebr...@gmail.com> wrote:
>>
>> Hi Sven -
>>
>> This is in Pharo 6.1.
>>
>> There is an API I'm using which sometimes returns a string only containing a
>> single instance of the number 0 in the "Expires" field, so the
>> #expiresTimeStamp method sends that to #parseHttpDate: and since it can't be
>> parsed into a date an error is thrown. Seems like Zinc should be able to
>> handle that based on the spec here:
>> https://tools.ietf.org/html/rfc7234#section-5.3
>>
>>
>> Would you rather I changed the implementation of parseHttpDate: to add an
>> empty check on the parsed tokens e.g.
>
> IIRC some months ago, there were some changes to this area, but I forgot the
> details and the before state.
>
> Anyway, reading RFC7234's section that you reference, maybe we do need to
> make another change, but I am not sure.
>
> Let's start by explaining what there is today and why it is like that:
> #parseHttpDate: and #expiresTimestamp do indeed throw exceptions (by design),
> while #isExpired is the more high level access method.
>
> Consider:
>
> (ZnCookie name: 'foo' value: '100') isExpired.
>
> => false
>
> ZnUtils parseHttpDate: '0'.
>
> => Boom
>
> (ZnCookie name: 'foo' value: '100') expires: '0'; expiresTimeStamp.
>
> => Boom
>
> (ZnCookie name: 'foo' value: '100') expires: '0'; isExpired.
>
> => false
>
> Could you use #isExpired and does it do what you want/expect ?
>
> Reading the spec again, I am not 100% sure about the error case (existing but
> zero or wrong Expires header), maybe that should be true instead of false.
>
> isExpired
> | expirationTimeStamp |
> (self hasAttribute: 'expire') ifFalse: [ ^ false ].
> [ expirationTimeStamp := self expiresTimeStamp ] on: Error do: [ ^
> false ].
> "note that max-age (#maxage) is not used"
> ^ expirationTimeStamp asUTC < DateAndTime now asUTC
>
> What do you think ?
> Anyone else with an opinion ?
I just noticed that we are mixing request/response http headers with cookies,
which might not be a good idea.
> Sven
>
>> parseHttpDate: string
>> "self parseHttpDate: 'Tue, 13 Sep 2011 08:04:49 GMT'."
>> "self parseHttpDate: 'Tue, 13-Sep-2011 08:04:51 GMT'."
>> "self parseHttpDate: 'Tue Jan 01 00:00:01 2036 GMT'."
>>
>> | tokens day month year hour minute second months map yearToken |
>> string size = 1 ifTrue:[ ^DateAndTime epoch ].
>> tokens := (string findTokens: #( $ $- $: $, )) allButFirst.
>>
>> ...
>>
>>
>> Or add a string size guard check in #expiresTimeStamp ? e.g.
>>
>> expiresTimeStamp
>> self expires
>> ifNil: [ ^ DateAndTime now + 1 day ]
>> ifNotNil: [ :exp |
>> exp size = 1
>> ifTrue: [ ^ DateAndTime epoch ].
>> ^ ZnUtils parseHttpDate: exp ]
>>
>>
>> Thanks
>>
>> Paul
