Re: [Jsonpath] Fwd: I-D Action: draft-bormann-jsonpath-iregexp-04.txt

[Glyn Normington <[e-mail]>]

I agree with your implementation cost argument.

Also, without a full test suite of all cases of syntactically invalid
strings (is such a thing even theoretically feasible?), implementations'
syntax checking would probably vary, impacting interoperation.

So, yes, I could live with a MAY.

On Mon, 25 Apr 2022, 18:26 Tim Bray, <[e-mail]> wrote:

> I can't disagree.  All I can add is that a MUST is a pretty heavy hammer.
> I think that for any of the well-known libraries, I could construct and
> ship an I-Regexp translator in a very short time, in some cases hours.
> Building a syntax checker would be a much more demanding task, and writing
> the test suite for such a checker would be very discouraging.
>
> While I at least partially agree with you, here's a counter-proposal: Put
> in a MAY (or SHOULD?)  and a paragraph containing basically the text of
> your email 5 minutes ago, encouraging implementors to do this.
>
> On Mon, Apr 25, 2022 at 10:07 AM Glyn Normington <
> [e-mail]> wrote:
>
>> If non-conforming regexps are handled, I think (at least perceived)
>> interoperation could suffer. Suppose someone gets used to using a
>> particular I-Regexp implementation and fails to notice they have stepped
>> outside the official syntax. Then at some point, they try another i-Regexp
>> implementation and either (a) their regexp syntax is rejected as invalid or
>> (b) the behaviour is different.
>>
>