Fw: Comments on LDP-Paging - 5.1.5 whole != sum(parts)

Moving SS reply on-list.

wrt union, all I'm saying is: since LDP Paging only defines how to combine 
pages when the paged resource is a LDP-RS, 'union' is not going to be 
necessarily the right operation for all LDPRs.  If I think about using 
this in Google Maps (each tile = 1 page), it's obviously *the wrong* 
operation no matter how you define union in any commonly understood way. 
Yet it seems like a perfectly sensible use of Paging, were those folks so 
inclined (disclaimer: I don't work for that company, don't know anyone in 
that dev group, and so on - purely speculative example on my part.)

I did verify that the current text is a straight copy from 5005 paged 
feeds.  Given that LDP Paging adds canonical etags as a way for clients to 
detect when the container changed, which paged feeds have no equivalent 
for, it is overly pessimistic now (but not when it was copied).  There's 
also a language mapping error (paged resource ~= logical feed, not paged 
feed) that we can fix given the updates I made earlier this week, wherein 
'page sequence' (which had crept into the draft via the email discussions) 
was promoted to a formally defined term.

Strawman revision:
5.1.5 LDP Paging clients SHOULD NOT treat a page sequence as equivalent to 
the paged resource when the _paged resource changed_ [link to 6.2.8] 
_during retrieval of the page sequence_ [link to 6.2.7].


I'm tempted to say Must Not, given the addition of the "changed" 
qualification vs 5005, but I can live with either.  Can anyone NOT live 
with Must Not?  Keep in mind this is a constraint on LDP Paging clients, 
not servers.



Best Regards, John

Voice US 845-435-9470  BluePages
Cloud and Smarter Infrastructure OSLC Lead

----- Forwarded by John Arwe/Poughkeepsie/IBM on 07/31/2014 10:35 AM -----

From:   Steve Speicher <sspeiche@gmail.com>
To:     John Arwe/Poughkeepsie/IBM@IBMUS
Date:   07/30/2014 04:19 PM
Subject:        Re: Comments on LDP-Paging - 5.1.5 whole != sum(parts)



On Tue, Jul 29, 2014 at 4:37 PM, John Arwe <johnarwe@us.ibm.com> wrote:
>
> > <#ldpp-client-paging-incomplete>
> > I find this normative statement hard to understand, at all:
> > [[5.1.5 LDP Paging clients SHOULD NOT present paged resources as
> > coherent or complete, or make assumptions to that effect.]]
> > I'm having a hard time with the words "present", "coherent" and
> > "complete" in this clause.
> >
> > I would think a clause such as this might make more sense:
> > [[5.1.5 LDP Paging clients SHOULD NOT present a union of all in-
> > sequence page resource representations as equivalent to the
> > representation for the paged resource.]]
>
> The proposed(?) update doesn't work for all LDPRs, does it?  It could be 
a useful example for LDP-RSs in particular, modulo using the more specific 
"graph union".
Why doesn't it work for all LDPRs, union in the proposed doesn't define it 
anything more than the normal mathematical term.

> I suspect the words that do so offend are carryovers from 5005, I'd have 
to check that.  "present" might be WebArch.
> "equivalent" ... 3986 chapter 5 -veined issues.
>
Hoping that "equivalent to the representation" would get what you were 
after.  So I would think equivalence would be defined by each of the 
representations.

If I were a client implementor, to comply I would have to present the 
paged resource as incoherent or incomplete.  An end user wouldn't like 
that and not use my client anymore.  Therefore I would probably just 
present it as "built from parts"; does this pass the compliance (SHOULD 
NOT) test of not-complete and not-coherent? I'm not sure.  The way I 
proposed to modify it, then yes I think I would understand I'm inline with 
the spec.

> At a higher level though, this was probably last visited before we had 
canonical etags, so apropos some of your other "this sounds scarier than 
it needs to" comments perhaps we could be more accurate.  
>
Yes, my reaction of "this sounds scarier  than it needs to be".  Though 
I'm losing how canonical etags softens it the way it was written.  Maybe 
that is not what you saying.

Have others weighed in?  Haven't seen anything on the list, I would like 
something "less scary" for LDP paging clients to have to present in order 
to conform.

Thanks,
Steve Speicher
http://stevespeicher.me


> [1] http://lists.w3.org/Archives/Public/public-ldp-wg/2014Jul/0059.html
>
> Best Regards, John
>
> Voice US 845-435-9470  BluePages
> Cloud and Smarter Infrastructure OSLC Lead
>