Re: recursive shapes in shape expressions

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



On 02/25/2015 01:39 AM, Iovka Boneva wrote:
> Hi,
> 
> Thank you Peter for carefully considering our semantics.
> 
> Here are comment on your last questions, some of the questions were
> answered in my previous mail.
> 
> 
> Le 24/02/2015 18:24, Peter F. Patel-Schneider a écrit :
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>> 
>> On 02/23/2015 07:25 AM, Iovka Boneva wrote:


[...]


>>>> However, sometimes it is not obvious what nodes have a particular 
>>>> recursive shape. For example, in ex:j1 :related ex:j1 . ex:j1
>>>> :status :Assigned . it is reasonable to consider ex:j1 as having
>>>> shape <IssueShape> but it is also reasonable to consider ex:j1 as
>>>> not having shape <IssueShape>. Does ex:j1 have shape <IssueShape>
>>>> here in Shape Expressions? Show how this can be determined in each
>>>> of the semantics for Shape Expressions.
>>> There is a valid typing that associates {<IssueShape>} with ex:j1. 
>>> Indeed, the constraints of <IssueShape> are satisfied in this node.
>>> As a side remark, I do not see why it is reasonable to consider that
>>> ex:j1 should not have <IssueShape>
>> Your semantics shows why this is reasonable.  In your semantics it is
>> be permissable to have ex:j1 not typed with <IssueShape> if there is
>> another type to which ex:j1 could belong.
> I'm still not sure to understand, but let me comment, maybe I answer the 
> question.
> 
> With the unique maximal type, we associate to every node all the types
> that it can have. In this particular case, ex:j1 satisfies the
> constraints from <IssueShape>, so it will have that type in the maximal
> typing. If we now consider starting with pre-typing of root nodes, then
> the aim is to ensure that the roots satisfy the required shapes, and in
> that case it might not be necessary to associate <IssueShape> with
> ex:j1.
> 
> Recall that typing is not unique, so indeed we can have valid typings in
> which ex:j1 has <IssueShape> and valid typings in which it has not.
> However, there is a unique maximal typing, and a unique minimal typing if
> a pre-typing is given. And that's really a nice property !
> 

Although it is sometimes nice to go to a maximal typing, it is not
self-evident that maximal typing is the correct way to go, even if unique
maximal typings exist.

peter
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBAgAGBQJU7griAAoJECjN6+QThfjzsrIIAK/6u1eXHBZJGPOy7pKB+mNm
kT7Xfiyip/RZVwOEa6mFHFNxlRVoFN8ULjmwJjYhB6wEtfWWdhGwKPuYgu4wzhHK
L66zM9+PtiRJiDRdCBfzWFzGIWi+BhYXr6KI1SSE4PbAtEgBOcxdskKKnwtZYLqR
P1cf1A+NRyGQXcb7iE6NThuRfvmvkYW7ia5FcS4osjfizGkQgMOnbsjpGH0WioSM
L1W10YI4q0YeEaNLJ1N2/5QrriOFJNRamGpCBybI7K9TrLgBYCuCYb20yNP4piU1
v1VSXygXaA5hL44bDtB+qrCuDOgC4SXD6BT301Wuk7OFTkwW79ErT638y40OHR8=
=c2vA
-----END PGP SIGNATURE-----

Received on Wednesday, 25 February 2015 17:48:48 UTC