LDP BP&G comments

I just committed a (very small, for me ;-) ) pile of typos and corrections 
that I think will be uncontroversial, at [1].  Cody's parallel changes 
were merged in, I spot checked.

Potentially controversial comments (not in [1])

1:  The order of pre-reqs in 1.3 seems odd.  RDFS before RDF Primer; LDP 
UCR last, Primer n-1 (should precede LDP itself for this audience, no?). 
No sign of Paging (which might be fine, if we have no BPGs) or LDPatch (no 
opinion yet, but given all the spirited past discussions seems likely).

2: In BPG 2.2, should we say anything about Link rel=type's relationship 
to rdf:type ?  Given some of the recent discussions around relationship 
type URIs/shortnames, we do seem to have at least rough consensus that the 
"set of types" for the resource, at least for the subset of cases where 
the effective HTTP request URI U of an LDP-RS appears in the 
representation as < U, rdf:type, t1...tn > triples, that the type set == 
the union of the Link rel=type and the < U , rdf:type , t1...tn > values. 
But maybe all those qualifications say it's below the radar; what do 
others think?

3: Title on 2.4 hurts eyes/ears.  I wrote down some alternatives quickly, 
none really much better.  Best I had was: Avoid dot-segments in URIs 
unless the base URI is known (the POST qualification then becomes part of 
the content).

4:  In 2.7 there's a whole set of "see also" links done in a completely 
different style than the other references, which looked kinda off.  You 
already have localBiblio entries, so I know it's not a "hth do you do that 
in respec" problem.

5: in 2.8  the "RDF does not define" statement  should be fixed unless 
Sandro (or another resident RDF 1.1 rep) thinks I'm reading [3] out of 
context.  Having re-checked [2] and [3], with [3] defining URIs for the 
Schema datatypes, and the explicit statement in [2] that it's *compatible* 
with XSD datatypes, any issue with [2] not defining the actual URIs seems 
to be past history.  Personally I think dateTimeStamp should be added, too 
(requiring a timezone, which is the only difference between that and 
dateTime, solves SO many implementation problems).  We might consider 
adding rdf:HTML too, although since we're not being limiting that one's 
more about optics than helping implementers.

6:  in 2.9, we talk about rdfs:range ... which of course is about 
inferencing [cue: allergic reactions].  I think we're also using it here 
not with an inferencing intent, but with something like a "proactive 
validation" intent - telling people what to do, instead of what to infer, 
kind of what "that other WG" is supposed to work on.  We should be clear 
as to our intent here, and then communicate that in a way unlikely to be 
confused ... if I'm right, omitting all mention of 'range' would be fine. 
It's more like "best practice object datatype".

7:  in 2.9, we mix some constraints from the base vocabulary definitions 
(dcterms:Agent on contributor) with ones we've added (description as XML 
literal, and with a subset of HTML tags allowed in the content).  We 
should be clear that we're adding our own constraints on top, and say 
that, even if we don't worry about being clear on what comes from where 
(follow your own nose if you don't like it our way).

8: (a nit I missed) the HTTP11 reference for qvalues should now be 
[[RFC7231]] ... if there are any lingering HTTP11 refs after that, poke me 
and I'll give you the right RFC for each; the 1:n mapping is not always 
obvious.




The biggest fish from [1]

NC1:  2.6 first few clauses reordered, and inserted the punch line.  In 
the original I had to read through another page (mostly 3 examples, with 
subtle differences) to figure out what the mysterious advantage is.  I 
also replaced "isomorphic" with "equivalent" somewhat reluctantly; 
isomorphic is more precise, but I fear the "...Guidelines" title is going 
to draw in relative newbies and that's a pretty scary word for such a 
crowd.  If someone replaces it (including back to the original), I won't 
object at all.

NC2:  in 2.7 I reworded the part about "assuming" base URIs [skin still 
crawling].  Might be good for the editors to look and be sure the more 
precise version isn't too scary for the audience.

NC3:  in 2.11 "primary URL" I added a line about trusting the server + 
7231 reference, ala Eric P's 2NN draft.



[1] https://dvcs.w3.org/hg/ldpwg/rev/d6b3d3929c79
[BPG] https://dvcs.w3.org/hg/ldpwg/raw-file/default/ldp-bp/ldp-bp.html
[2] http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/#section-Datatypes
[3] 
http://www.w3.org/TR/2014/REC-rdf11-concepts-20140225/#section-Datatypes

Best Regards, John

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