Re: GRDDL and OWL/XML

On May 13, 2008, at 3:59 PM, Dan Connolly wrote:
[snip]
> Yes, I remember, but I am not persuaded. Reference
> implementations are a good thing when we can manage them.

I would hazard that this perspective is not widely shared by the  
membership of the W3C. I've often heard this presented  
*axiomatically* (i.e., as a basic principle). Furthermore, I'm really  
skeptical that the W3C has the resources to be the steward of  
reference implementations...esp. when having a *comprehensive* test  
suite is (correctly) seen as far beyond the typical resources of a  
working group.

In any case, I think that that in a standards body, the argument  
*for* reference implementations (esp. when that's not part of the  
culture) is the one that needs support, not the contrary. Thus, it is  
you who must persuade me :)

The member organizations I've been connected, who generally do the  
very hard work of not just  widely used, open source implementations,  
not just in CR, but also tracking ongoing working group work, fine  
the idea of reference implementation at the W3C, esp. as a norm or  
default, unhappy making. It makes them uncomfortable with the W3C.

(I'm not speaking just for me...I've asked around quite a bit.) They  
are not alone with this.

We can debate about whether we're right to feel th at way, but I  
think that the W3C should do a better job of communicating whether  
this is a general policy, how it works, how it shows up in charter,  
is this only GRDDL.

This raises the risk of making GRDDL into something to be resisted  
everywhere :(

I provide this merely as data for you. It's up to you to decide what  
to do with that data. Evidently, you don't find what I've already  
said worth engaging in detail, so I guess we're at an impasse there.

>>> then it shouldn't use GRDDL.
>>
>> But I don't think GRDDL requires a reference implementation, nor is a
>> reference implementation a requirement for the utility of GRDDL
>
> I don't see much utility in using GRDDL if there isn't
> an XSLT transformation available online.

I've addressed this elsewhere. But if this is the case and consensus  
of the GRDDL community and WG (esp. if y'all tighten the spec) then  
there's nothing left for me to do than systematically oppose GRDDL,  
esp. wrt W3C specs.

I *don't want to go there*. Really and truly. I supported GRDDL in  
the OWL charter because I thought it would be useful to spec a GRDDL  
transformation (see my other reply).

(I'll point out that it's not just a reference implementation but a  
*default*, *automatically used* implementation. That raises the  
stakes for me.)

> I think it's fine for GRDDL-aware agents to optimize
> the transformation by having better implementations
> installed locally, but without the bootstrap
> implementation in the web, why bother with GRDDL
> markup at all?

Because I don't think the primary value of a GRDDL agent is in  
automatic extensibility, but in being the go to transformation agent?
[snip]

>> People are reading the SHOULD as a MUST.
>
> Really? I am not.

I'm not saying you are. That's "People" as in "some people" not "all  
people". To be clear, people in the OWL WG claimed that we would  
violate our charter and not conform with the GRDDL specification if  
we did not supply an auto-downloaded XSLT. Thus, we ended up with  
MUST talk instead of a debate on the merits.

This seems to be a problem with the GRDDL spec/documentation.

> I just haven't seen a good reason to make
> an exception in this case.
[snip]

I, of course, believe I've provided many excellent reasons. Thus far  
I've seen *no* substantive, argued out case for it.

I would think that, since presumably we're trying to gain consensus,  
and consensus is all about taking everyone's interests seriously,  
that my expressed interests don't give you pause enough to build a  
non-pure-assertional/rhetorical-question-based case. That's your  
right, but I trust you understand why I don't find that convincing! :)

It seems to me that for some people, it's such a no brainer that  
having autoloaded executed transformed is good there's no point to  
doing anything else (except, maybe, resource constraints). So, how do  
you talk to people (like me) who 1) don't think that it's a no  
brainer that autoloaded executed transforms* are a good idea (indeed,  
default against them) and 2) think that GRDDL can have value  
otherwise? I know that there are more people who fall into 1 and I  
also know that many of those lean heavily against 2 or lean against 2  
to the degree that it requires autoloaded executable transforms.

Thus, I think that it would help GRDDL's profile and adoption if the  
GRDDL community developed a better message for dealing with such people.

Cheers,
Bijan.

*This is an approximation because there are issues about multiple  
specifications, etc. etc.

Received on Tuesday, 13 May 2008 15:51:34 UTC