RE: [xri] XRD feedback from Scott

[Eran Hammer-Lahav <eran@hueniverse.com>]

I am not sure how much we need to spell out the relation between the transport caching directives to the document's expiration directives. While they are used pretty much the same way, they have slightly different semantic meanings. In practice, if used over HTTP, the way the spec is written today is how it should be implemented (the earliest expiration date of the two). Otherwise you violate either HTTP or XRD (or both).

Maybe we should simply say that clients should comply with the caching rules of the transport used and leave it at that.

EHL

> -----Original Message-----
> From: Joseph Anthony Pasquale Holsten
> [mailto:joseph.holsten@cordance.net]
> Sent: Thursday, May 28, 2009 3:28 AM
> To: Scott Cantor
> Cc: 'XRI TC'
> Subject: Re: [xri] XRD feedback from Scott
> 
> Scott Cantor supposedly wrote:
> > Joseph Anthony Pasquale Holsten wrote on 2009-05-27:
> >> Though that sounds confusingly similar to RDF.
> > I did, umm, ask that question...doesn't RDF already do this kind of
> > stuff?
> 
> Sure, but RDF's quite a bit more powerful and complicated. You can
> also build theorem provers, relational DBs, &c with it. It's just not
> very easy to use.
> 
> For example, a project similar to XRD is POWDER. While in XRD you
> query with a relation to get a set of URIs (I've got a phone, who can
> I call?), in POWDER you query with a URI to get of relations (I need
> someone named Smith, can I call one?).  They wanted to have an RDF
> notation so they can infer from the POWDER record. I wouldn't put
> POWDER against XRD in a beauty contest, but compare regular POWDER
> against the generated RDF in their Formal Semantics spec sometime.
> POWDER is a few orders of magnitude easier to read than POWDER-S
> because they built a high-level language in XML and transform it into
> low level RDF.
> 
> The trouble is trying to store everything in primitive data structures
> without providing any powerful abstraction to deal with it. Kind of
> like trying to use lisp without macros, lambdas, or eval. Unless
> you've got nothing better to do, I'd advise against it.
> 
> I've been working on an informal vocab to transform XRD into RDF, but
> it's not very fun to use either.
> 
> >> [...] why not say Expires: MUST match or MUST NOT be present. Seems
> >> like there might be caching considerations too.
> >
> > Yes, that's a good point, but by saying it has to match, you end up
> > creating a condition that has to be tested in the implementation, so
> > I think it's simpler just to ignore the HTTP layer there.
> 
> Sounds fine. But if it's going to be ignored, then they SHOULD NOT set
> Expires: or anything else we have to ignore. Otherwise someone's going
> to tear their hear out while debugging, trying to understand what
> idiot explicitly broke compat w/ HTTP, until they find just that line
> in XRD.
> 
> Joseph Holsten
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe from this mail list, you must leave the OASIS TC that
> generates this mail.  Follow this link to all your TCs in OASIS at:
> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php