Re: Updated Conformance and Test Cases

On 12 Nov 2008, at 15:33, Ian Horrocks wrote:

> On 11 Nov 2008, at 23:21, Alan Ruttenberg wrote:
>
>> On Tue, Nov 11, 2008 at 2:00 PM, Ian Horrocks
>> <ian.horrocks@comlab.ox.ac.uk> wrote:
>>> On 5 Nov 2008, at 17:47, Alan Ruttenberg wrote:
[snip]
>>>> "must return Error if an input document uses datatypes that are not
>>>> supported by its datatype map or literals that it does not support
>>>> (for example, very large integers); and"
>>>> I wonder whether the appropriate response here is Unknown rather  
>>>> than
>>>> Error. It seems rather like not having enough resources to evaluate
>>>> the check.
>>>
>>> Error seems right here; Error is also returned "if the  
>>> computation fails,
>>> for example as a result of exceeding resource limits".
>>
>> This doesn't seem to be a case of exceeding resource limits. It's a
>> case of making a choice to not supporting what is otherwise valid  
>> OWL.
>>  If tool claimed to support arbitrary precision integers and then
>> failed because it ran out of memory or when processing a million  
>> digit
>> integer then I would consider it a resource failure. If it doesn't
>> even try then I think it's clearly an unknown.
>
> Well, I would say that Error applies to any failure at the parsing  
> stage. However, I would welcome opinions from other WG members.

+1

Also, if you fix the precision that you handle then you are fixing a  
resource limit. If something exceeds that limit (i.e., by being too  
long an integer) it exceeds a resource limit.

Some systems do this with e.g., recursion depth. Python had a limit  
of like 32 calls deep. Seems like a resource limit tome.

If my program looked at a file, saw that it was 10000megs and said,  
"No way I can handle something that big, so I won't even try" that  
seems to be a resource limit thing as well.

I wonder why we're wordsmithing at this level. I think "Error, I  
don't handle this datatype" or "Error, I don't handle numbers that  
big" is fine.

Cheers,
Bijan.

Received on Wednesday, 12 November 2008 15:53:53 UTC