Potentially redundant context rules

[Patrik Fältström]

Can people that know bidi inside and out assure me that the rule(s) in  
bidi already today, without any changes, implement exactly the same  
rule as tables context rule does?

That claim I have missed, as I thought some changes where needed.

    Patrik

On 28 jul 2009, at 18.05, Chris Wright <chris at ausregistry.com.au> wrote:

> Patrik et al.
>
> My reading of that thread is as follows (my extremely short  
> summarised opinion only, focusing only on hyphen and Arabic digits  
> context rules as these are the only two that I was discussing):
>
>> Mark Davis: Existing context rules shouldn't exist
>
>> Andrew Sullivan: (about hyphens) If it is a requirement of the DNS  
>> then surely it should be a rule
> [to which I would say, DNS imposes no requirements on U-labels, this  
> working group would come up with those requirements, if we are to  
> make this rule because we wish to have U-labels also conform to the  
> host name rules for hyphens then surely a context rule that can be  
> ignored is not sufficient and that making it a protocol rule would  
> be better as is done with the consecutive hyphens]
>
>> Mark Davis: (about hyphens) Confused about it being a requirement  
>> of the DNS, however, makes the same point that we do about it  
>> should be a protocol rule
>
>> Alireza Saleh: (about Arabic numbers) Makes a statement that he  
>> wouldn't like to see the Arabic number restrictions as either  
>> context rules OR BIDI rules!
>
>> Michel Suignard: (about hyphens) Also got confused about the hyphen  
>> rule being a requirement of the DNS, (about Arabic numerals)  
>> supported having the Arabic numeral context rules in BIDI only  
>> (well said it "is probably fine")
>
>> Andrew Sullivan: (about hyphens )Said he doesn't care, but  
>> preference is a context rule, with no reasoning as to why he  
>> prefers a context rule
>
>> John Klensin: (about Arabic numbers) Supported the rules about  
>> Arabic being in BIDI, said they ended up in context so he didn't  
>> argue, but he preferred them in BIDI
> [and our point is that the rules are in BOTH context and BIDI at the  
> moment, thus redundant and only NEED to be in one!]
> (about hyphens) He then went on to say that the hyphens should not  
> be protocol but context, because DNS is meant to be about octets
> [following that argument then I don't understand why the "double  
> hyphen" rule gets to make it into the protocol and not be a context  
> rule as well? I think John may have confused the order in which he  
> discussed the double hyphen and leading and trailing hyphens but I  
> am not sure]
>
>> Mark Davis: BIDI is good, doesn't like context rules, but doesn't  
>> really care as he will just ignore them
>
> Now the only two issues we are raising is the following:
>
> The Arabic number rule is enforced twice, once in BIDI and once in  
> context rules, we are not saying it shouldn't be enforced (ie we are  
> not saying ignore the advice of the Arabic experts), we are saying  
> it only needs to be enforced once in one place! It is redundant to  
> do it twice! And then we are saying that our preference is to see it  
> remain in BIDI and have the context rule version taken away. <--- I  
> haven't seen anyone comment on this, or alternatively explain why  
> the two versions are actually required
>
> The other argument about hyphen is again one of consistency, if we  
> can enforce '--' in protocol then why can't we also enforce the  
> leading and trailing dash in protocol as well (where it can't be  
> ignored), if the intention is to give u-labels similar rules to the  
> hostname rules. <---- I haven't seen anyone explain why the one  
> exception is allowed but the other is not
>
> Thanks
>
> Chris
>
>
> -----Original Message-----
> From: Patrik Fältström [mailto:patrik at frobbit.se]
> Sent: Wednesday, 29 July 2009 1:21 AM
> To: Chris Wright
> Cc: idna-update at alvestrand.no
> Subject: Re: Potentially redundant context rules
>
> See the latest discussion with subject "CONTEXTO Proposal".
>
> I do not draw the same conclusions as you based on the latest thread.
>
> What do others think?
>
>    Patrik
>
> On 28 jul 2009, at 15.50, Chris Wright wrote:
>
>> Patrik,
>>
>> As per our conversation just after the working group meeting
>> yesterday, this is a reminder to look into why there are context  
>> rules
>> for Arabic digits in the tables document (the last two context
>> rules) when these tests are already covered by the rules in the BIDI
>> document (i.e. they are redundant). If there was some reason for this
>> (none that we seem to be able to find) we would suggest that this
>> rational be added in the appropriate place in the rational / tables/
>> or BIDI document, otherwise the redundant context rules should be
>> removed and the code points returned to PVALID status.
>>
>> Some previous discussions on the matter seem to support that this was
>> to be done in the BIDI document OR as context rules eg.
>>
>> http://www.alvestrand.no/pipermail/idna-update/2008-November/ 
>> 003024.ht
>> ml
>>
>> It appears that, perhaps for different reasons, it has actually ended
>> up being done in both places. Either way, even if each set of rules
>> exist for different reasons, the context rules are redundant.
>>
>> Thanks
>>
>> Chris
>>
>>
>
> _______________________________________________
> Idna-update mailing list
> Idna-update at alvestrand.no
> http://www.alvestrand.no/mailman/listinfo/idna-update
>