Actually, how about just the character position in the input stream?  That
would avoid any complexities of different line-ending styles.  It could
simply be named "position".


On Sun, Oct 13, 2013 at 9:14 PM, Y. Richard Yang <[email protected]> wrote:

> I am fine with adding two optional fields. Regarding an optional
> "character" field, the suggested semantics is the character position at the
> line. To avoid ambiguity, how about name them "line-number" and
> "character-number" so that their JSONNumber type is not surprising?
>
> Richard
>
>
> On Sat, Oct 12, 2013 at 1:43 PM, Richard Alimi <[email protected]> wrote:
>
>> We used to have a 'reason' field, but it was suggested in the the
>> apps-area review that it be removed because of the concern over
>> internationalization.  Given that it's WGLC I would prefer not to revisit
>> that.
>>
>> I don't have a problem with adding optional line and character fields,
>> given that they can be conveyed as integers.  I don't think it should be
>> required, since not all parsers may provide that information.
>>
>> Rich
>>
>>
>> On Fri, Oct 11, 2013 at 5:54 PM, Y. Richard Yang <[email protected]> wrote:
>>
>>> Hi Wendy,
>>>
>>> Interesting suggestion. I assume that the fields will be optional,
>>> right? A consideration is that the syntax errors typically are caught by a
>>> JSON parser. I know that you know quite a few parsers. Will the existing
>>> parsers give such information (e.g., line, character)?
>>>
>>> To others: if you have used multiple parsers, your experience can be
>>> quite helpful here.
>>>
>>> Thanks!
>>>
>>> Richard
>>>
>>>
>>> On Thu, Oct 10, 2013 at 12:50 PM, Wendy Roome <
>>> [email protected]> wrote:
>>>
>>>> As current;y defined, E_SYNTAX gives no indication as to where the error
>>>> is or what was wrong. So how about adding the following additional
>>>> fields
>>>> for E_SYNTAX:
>>>>
>>>> "line"  Optional line number (first line is 1)
>>>> "character"  Optional character in line (first character is 1)
>>>> "reason"  Optional error message, which may or may not include line and
>>>> character.
>>>>
>>>> This raises the question of what language to use for "reason". At the
>>>> risk
>>>> of coming across as an English chauvinist, I suggest that it be in
>>>> English. After all, the RFC is in English, and the field names are
>>>> English. So it's pretty unlikely that server and client programmers
>>>> won't
>>>> understand enough English to deal with a simple error message. And I
>>>> don't
>>>> see a client library displaying that message to the end user; instead
>>>> the
>>>> lib would log it for analysis by the developers.
>>>>
>>>> And if a server isn't able to give an English error message, use "line"
>>>> and "character".
>>>>
>>>>         - Wendy Roome
>>>>
>>>>
>>>> _______________________________________________
>>>> alto mailing list
>>>> [email protected]
>>>> https://www.ietf.org/mailman/listinfo/alto
>>>>
>>>
>>>
>>> _______________________________________________
>>> alto mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/alto
>>>
>>>
>>
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to