Linked Data Platform (LDP) Working Group Teleconference

10 Mar 2014

See also: IRC log


Arnaud, JohnArwe, Sandro, codyburleson, TallTed, Roger, SteveS, deiu, Ashok, nmihindu, ericP


<trackbot> Date: 10 March 2014

<scribe> scribenick: Deiu

scribenick Deiu

<scribe> Scribe: Deiu

approval of last minutes

<Arnaud> http://www.w3.org/2013/meeting/ldp/2014-03-03

Arnaud: minutes approved since there were no objections
... any problem with next week's meeting? Same time, same hour difference

Actions and Issues

Arnaud: I used to track issues during the call, but I dropped it for a while. Now we can go back to tracking issues.
... re. Actions, we're now done with the spec, so we have to get back to the other deliverables
... there are some actions that need attention, any open actions? Any progress on them? (only 7 open but still)
... is the spec supposed to be published tomorrow

SteveS: everything is good to go for tomorrow

Arnaud: it could have been April 1st, but we thought we should avoid that particular date
... we have 3 weeks at the beginning of April to finish the CR
... we don't have time to finish the spec before the group expires
... but we're pretty much done
... we can close shop at the beginning of June; it won't affect the spec too much
... we can get an extension though, in case we want to respond to comments
... if people are interested in continuing working on LDP, we still have a "wish list", so we can work on a new charter
... that will take some time to be approved, so maybe we can get an extension for the current group to try and deal with some of the points on that list
... we need to agree on the exit critera for CR
... at least 2-3 implementations
... it may take some time for people to implement the spec, after the latest changes

sandro: we need implementations that are interoperable, and which is often done with a test suite

Arnaud: we should aim for demonstrating interop as much as possible, so the test suite is a priority

Use Cases and Requirements

Arnaud: there were improvements to the document, but we never got around to publishing it

<Arnaud> PROPOSED: Publish the latest UC&R draft

<SteveS> +1

<Ashok> +1

<roger> +1

<TallTed> +1

SteveS: I thought stevebattle15 started the process and he requested a date


<codyburleson> +1

<nmihindu> +1

<MiguelAraCo> +1

<Arnaud> RESOLVED: Publish the latest UC&R draft

<JohnArwe> +1

Arnaud: I will let everyone know when it's published


Arnaud: we had talked about having a 3 day meeting, and maybe reduce it to 2 days

<SteveS> maybe 3rd day can be test suite and other hacking

Arnaud: I want to confirm that this will be a 3 day meeting (around 15-17 April)

<sandro> +1 3 day meeting, especially if we can be doing interop testing

Arnaud: there is a list of items to discuss

+1 for hacking

<codyburleson> +1 3 days (or more)

scribe: we have two specs now (LDP + Pagind&Ordering), so we have to make sure that they move forward
... we also have PATCH details to discuss
... we can spend some more time to discuss PATCH format, if we don't have too many LC comments

<Arnaud> PROPOSED: Confirm 3 day meeting: 15-17 April


<SteveS> +1

<MiguelAraCo> +1

<roger> +1

<TallTed> +1

Arnaud: people can go to the page and indicate there if they want to participate to the 3 day f2f

<sandro> +1

<JohnArwe> +1

<Arnaud> RESOLVED: Confirm 3 day meeting: 15-17 April

<codyburleson> That might be good, Sandro - if anybody cared to stay an extra day for extra "hackathon"

Arnaud: so now back to paging
... based on feedback from sandro, we have agreed to separate paging from the main LDP spec
... paging and ordering are now in a new draft
... we need to make progress, following the discussions on the mailing list
... there is currently no stability
... how much can the client miss when walking the pages
... currently, the spec does not guarantee anything
... sandro made the point that we need more guarantees
... we need to discuss this further
... I want to make sure that since this is an important aspect of LDP, we should not forget about it

<sandro> deiu, we hope you're dialing back

<ericP> scribenick: ericP

sandro: to what extent will paging work for arbitrary graphs (vs. ldp:Containers)

<deiu> Zakim: TimBL is me

sandro: we always had it mind that it would work in LDPRs, but we never worked through e.g. ordering

<deiu> Arnaud: member submission was attached to containers

scribenick deiu

<SteveS> FYI I created a new tracker "Product" for "LDP Paging Spec" https://www.w3.org/2012/ldp/track/products/8

<JohnArwe> Note: internally, we had paging for everything. it was easier to explain for containers only so we moved it there.

<deiu> sandro: paging makes sense if you have ordering, and ordering was not defined for arbitrary LDPRs

<deiu> Arnaud: should we open issues to start keeping track of these items more efficiently

<deiu> ... JohnArwe do you have any input?

<deiu> JohnArwe: we had paging for arbitrary LDPRs initially, but we moved it over to containers since it was easier to explain to people

<deiu> ... pages came back and the implementation would manage the rest

<deiu> sandro: you would order on the properties of the subject

<deiu> ... group by subject and order by a property (say foaf:name)

<deiu> ... for subjects that _have_ a foaf:name

<deiu> ... if you don't order the graph, then paging is just random sampling

<deiu> @@@: none of the serializers do any ordering now

<deiu> sandro: this would pose a problem for very large files

<deiu> ... there is also filtering (vs paging)

<deiu> Zakim: who is talking?

<Arnaud> john

<Arnaud> JohnArwe:

<deiu> JohnArwe: we weren't thinking about it in terms of a function that assigns resources to a page

<deiu> sandro: you want pages to be a fixed (or approx) byte size

<deiu> ... request more pages as you need them

<deiu> JohnArwe: you also want to be able to skip pages and/or support actions like "Home" or "Back" in the browser

<deiu> Arnaud: ok, so we need to capture the issues

<deiu> ... one: how lossy do we allow paging to be?

<deiu> ... two: how do we do paging if there is no ordering, since ordering is defined for LDPCs

<deiu> ... paging is defined for LDPRs

<deiu> ... we can argue about its usefulness

<deiu> JohnArwe: your client may not be only an LDP client, and so the client may know more (has support for extra features)

<deiu> ... it doesn't mean that is a problem if it isn't in the LDP spec

<deiu> Arnaud: because we don't have a solution now, it doesn't mean that we won't find one later

<deiu> sandro: is paging only used for flow control or does it allow for application grouping?

<Zakim> SteveS, you wanted to ask of there is any problem with paging spec just reusing LDP namespace


<JohnArwe> I think Steve is asking if the paging spec-defined elements get a different NSURI

<deiu> SteveS: we have some terms in the LDP namespace for paging

<deiu> ... do extensions go into the LDP namespace?

downside is that it's difficult to assert the stability policy

<deiu> sandro: you'd like people to play with extensions in a different namespace, but you'd also want to keep everything in the same space

<deiu> Arnaud: with the spec the way it is, sandro, do you have a problem with the way things are now, or is it ok?

<deiu> ... if we don't define ordering, can we keep the paging in LDPR?

<deiu> sandro: I agree with JohnArwe about adding ordering later, but I'm not sure about how we will implement it

<deiu> JohnArwe: how do you talk about the next page when you don't know how to sort it?

<deiu> sandro: the server must figure out some kind or ordering so that it will not repeat things nor leave things out

<deiu> Arnaud: my point is that if in your implementation you don't care about paging, you can choose not to do paging on graphs, for which you don't have a defined order of pages

<deiu> sandro: the paging spec is optional, I agree

<deiu> Arnaud: if you say you do paging, you have to be compliant and do paging on all resource types

<deiu> sandro: you can ignore the requests to do paging

<deiu> ... the paging will be initiated by the client, say "Preferred page size: 10MB"

<deiu> Arnaud: the server is in charge of deciding what resources get paged

<deiu> sandro: we need to come up with an arbitrary way of doing paging

<deiu> ... so that if the server can do paging for one resource type, it will be able to do it for others too

<deiu> JohnArwe: if I have a small resource (1-2 triples), am I required to break those into multiple pages (by supporting paging)?

<deiu> sandro: if the client says "Max page size: 10MB" then the server will not send more than 10MB

<deiu> ... if a triple pushes you over the page size, you'll get another page

<deiu> Arnaud: I want to raise 2 issues

<deiu> ... 1) how lossy is paging?

<JohnArwe> The server-side issue with the size limit is that the size is not so easy to predict (with precision) when using OTS serializers.

<deiu> ... 2) what does it mean to have paging without ordering for LDPRs?

<sandro> sandro: the exception being literals over the page size -- the server still needs to send those as is

<deiu> sandro: the issue of how to page on LDPR is a different issue

<deiu> sandro: sorry, how to order?

<deiu> ... ordering arbitrary LDPRs

<JohnArwe> When you're near the page size, as a server you have to guess what fits, and when you're wrong you see a spike in overhead as you retry (remove a bit, re-serialize, test again), in pathological cases repeatedly

<SteveS> FYI I created a new tracker "Product" for "LDP Paging Spec" https://www.w3.org/2012/ldp/track/products/8

<deiu> Arnaud: we need to make sure that raised issues match the right spec

<deiu> Arnaud: we need to get back on progress on deliverables: test suite, etc.

<nmihindu> I can ping Raul and ask about that

<deiu> ... we also have the ACL document; best practices and guidelines

<deiu> ... it might be good to take another look at the primer to check that is in sync

<deiu> ... we need to identify how we can make progress

<codyburleson> After the last major update to the spec, I am guessing that all the docs need review and revision again. The last changes in spec were pretty dramatic.

<SteveS> sandro, erice, deiu, any idea why mercurial is so slow ?

<SteveS> ..much slower last couple of weeks

<deiu> SteveS: no idea really, you're the first person I know that complained about it

<Arnaud> trackbot, end meeting

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2014-03-10 15:01:42 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.138  of Date: 2013-04-25 13:59:11  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/for that we need a test suite/which is often done with a test suite/
Found ScribeNick: Deiu
Found Scribe: Deiu
Inferring ScribeNick: deiu
Found ScribeNick: ericP
ScribeNicks: Deiu, ericP
Default Present: Arnaud, JohnArwe, Sandro, codyburleson, TallTed, Roger, SteveS, deiu, Ashok, nmihindu, ericP
Present: Arnaud JohnArwe Sandro codyburleson TallTed Roger SteveS deiu Ashok nmihindu ericP

WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth

Found Date: 10 Mar 2014
Guessing minutes URL: http://www.w3.org/2014/03/10-ldp-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]