| [higgins-dev] Notes for Higgins dev call - June 10, 2010 |
|
Notes for Higgins dev call – June 10, 2010 Attendees •
Valery Kokhan - Azigo Ukraine •
Mike McIntosh - Azigo •
Mary Ruddy - Meristic •
Drummond Reed - Cordance •
Paul Trevithick - Azigo •
Hank Mauldin - Cisco •
Alexander Yuhimenko •
Sergey Lyakhov •
Joseph Boyle LOGISTICS: •
Time: noon Eastern (16:00 UTC) U.S. •
Dial-in: 1-201-793-9022 passcode 7990866# AGENDA (Discussion of wiki updates) 1) [Paul] HIGGINS 1.1 RELEASE PLANNING •
Project plan [1] is this aligned with project metadata? •
Higgins 1.1 release milestone plan [2] - walk through
each solution and discuss •
[Paul] I updated the menu. On the home page
for Higgins, I got rid of our own custom plan. It now takes you to the
project plan page. For Higgins 1.1, I projected out the dates and
created a new M9. There is a link for each milestone. I’m
planning on removing the dates from the individual wiki subpages to
remove redundancy. I’ve updated the description at the top. I
will add a link for the plan for 2.0 here. Can only have one formal plan
at a time. •
[Paul] On the Higgins 2.0 plan (temporary wiki)
I’ve created 4 subpages for sub milestones. There isn’t much detail
there yet, but it is the genesis of the planning. Maybe by tomorrow the
wiki pages links will be fixed. •
[Paul] Mary this is the official eclipse format.
Has the same problem that when you go there you are lost as the links are for
Eclipse overall rather than for the Higgins project. •
[Paul] I think is it far more important for us to have
a single plan that we update in the official format. So we don’t
have discrepancies between various plan versions. •
[Paul] Dates for 2.0 are just estimates for now. •
[Paul] According to this plan we are working on
1.1M8. The only items not done are more formatting enhancements for the
down load page, assigned to Valery. Is that still a part of the plan for
next week, or should that be moved to another milestone? •
[Valery] This is actually done. I believe it is done. •
[Paul] There is also an item for Mary •
[Mary] It is not large. I just need to get around to
it. I’m not traveling for the rest of the week. •
[Mike] 290669 is finished. 334… implement
new build result pages will be done this week. •
[Paul] We should focus on taking the items that are
part of the 1.1 wiki pages and slotting them into milestone M9. Right now
there is nothing there as it is a brand new milestone. Alex, you were
mentioning the card sync item. •
[Alex] Yes. •
[Paul] For milestone M9, I could create a copy of
the 1.1 page, all its sections, because there is no place to take notes on what
you just said. •
[Paul] Hit refresh, there is now a section for each
solution on the M9 pages. Alex, are you saying that for 1.1M9 we should fix the
card sync deployment instructions? •
[Alex] Yes. •
[Paul] OK. I have added an item for that. So we
now have a template for deciding what we will do for M9. •
[Mike] I will do this offline before next week’s
call. •
[Paul] The 1.1 page is really a list of question to ask
for 1.1. Next task is to have a list of bugzilla tickets. 2) [Paul] IDAS INTERFACE CHANGES FOR H2.0 •
Remove IHasAttributes on IAttributes (no harm in doing
this for Higgins 1.1 as well) •
Sub-contexts: what is our plan - are we going to
implement sub-contexts using NG4Jena? •
Graph traversal: which are we going to implement? o
Walk a link to an entity in a different context
and you get back the UDI or o
Walk a link to any entity and get back either
the real thing or a lazily loaded proxy o
Some other idea •
Other changes? •
[Paul] Is Sergey on the phone.? •
[Sergey] Yes. •
[Paul] Was trying to make a list all in one place. On
the Higgins 2.0 page, were there any items that I missed? •
[Paul] At the bottom of the page is a building blocks
section. One is IDAS interface changes. I wanted to first make a
list. The first item is removing iHasAttributes in iAttibutes. I
thought we had agreed on making that change. I want to go ahead and get it done
with. I don’t think anything uses it. •
[Sergey] I will remove. •
[Paul] Should it be 1.1M8 or M9? •
[Sergey] I would prefer 1.1M9. •
[Paul] I will update that now. Hit refresh. •
[Paul] Subcontexts. If this is going to be a huge
project, we should reconsider doing it and just go with putting metadata like
entities in the main graph. On the other hand, if you feel that … I
would like to do it… from the email last week. It looked like there were
a lot of problems. I want to understand. •
[Sergey] Yes, we can implement it, but in my opinion,
If we need to put meta data… •
[Valery] Actually Paul, other approaches could have
same problem. •
[Paul] So what you are saying is don’t bother
with subcontexts, just have a meta data convention. •
[Valery] Yes. •
[Paul] So, the RDF community is using named graphs, but
the named graph for the JENA implementation is so buggy and problematic that
since I think we will need to move to another system anyway, if this
isn’t fairly easy to do and will be buggy, it is not such a good
idea. So basically we aren’t doing it. So we will just develop our
own conventions and change our client code to ignore or find those
things… Mike, are we in agreement on all this? •
[Mike] Yes. •
[Sergey] I would prefer to not put this code in the
client code. •
[Paul] So you would change the IdAS interface so that
the client of IdAS can say, give me the attributes about an attribute? •
[Serge] It is not exactly the same interface we just
deleted. •
[Paul] So we need a proposal for what we do. •
[Valery] Let me make a proposal with Sergey. Then we
can discuss. •
[Paul] So stop work on subcontexts, and work on a
design. •
[Paul] The next topic is graph transversal. So we
can walk a graph within a contest or a logical graph that spans contexts.
My preference has always been that it would be nice if the client didn’t
have to always be aware of context boundaries. •
[Paul] What I was thinking is that it would throw an
exception if it couldn’t get access. It would simplify the client
code from having to text every time. What do you think? •
[Mike] Is it any more trouble to test, then to attempt
and have an exception handler? It doesn’t seem like it is any more
or less code. •
[Paul] I haven’t looked at info grid. It is
the only other thing I know of that handles this to see how it handles
this. •
[Paul] This seems like an area where we need a design
and discussion. It reminds me of a third item. Updating the model
interfaces and hiding the model context. Behind those interfaces. •
[Sergey] I’m pretty sure that is the way to go. •
[Paul] These are things that we want to get done fairly
soon. •
[Mike] If the API can handle the exception locally,
without messing up the client code, I’m all for it. Just need to
figure out how to make it work. 3) [Paul] XDI HARMONIZATION (new call at 1pm Eastern) •
Discuss the idea of having weekly 30 minute calls
immediately after the regular Higgins call that ends at 1pm Eastern. •
The idea is to work on harmonization between the
Higgins flavor of XDI and a new flavor now under development by the XDI
community. •
Each week we'd try to chip away. •
[Paul] The last time, I talked to Drummond about
this issue, about the difference between Higgins graph model and XDI.
I’m developing some new ideas. And there is some interest.
People beyond Higgins are looking to use XDI in the pure sense, and since we
are not really (fully)interoperable with XDI, it creates an issue. It is
not an immediate problem, but Markus is developing his own XDI data store, and
other commercial companies are exploring developing XDI data stores. The
idea would be to allow our active client to read from a Higgins data store and
other non Higgins based data stores. In other words, have the integration
happen at a client. And that only occurs of we are all using the same
messages. We are 80% or maybe 50% compatible. The thought is that we
could have a conversion every week for 30 minutes. I need to create a
wiki page with all these issues. So that in a few minutes we can talk
about it. •
[Paul] Mary I’ m done (with the topics for the
regular Higgins call). •
[Paul]The XDI call is on the same call bridge. It is a
continuation of the Higgins call with a fixed topic and fixed time. •
[End basic Higgins call] •
[Paul] Resuming with the XDI specific call. •
This is the call with the XDI topic. Have just
created this wiki page [3] and hope it represents the latest thinking. •
[Paul] Let’s go through the wiki page. It is the
result of conversations in Munich. •
[Paul] Higgins speaks an XDI dialect. Hope to over time
adjust the Higgins code so that it is conformant with PDX (which will be an XDI
dialect that is maintained at XDI .org.) •
[Drummond] There is pure XDI. PDX is just a
dialect. Joe is that clear? •
[Joe] Yes. •
[Paul] The next item refers to CDM/PDM. Taken
together they define the types of graphs that will be found in production
Higgins space systems. •
[Drummond] I suggest the term Higgins RDF graph. •
[Paul] Higgins is usually more general that RDF, and
occasionally more restrictive. We can also use XDI’s for
URI’s example. •
[Paul] We have 4 kinds of issues to resolve. So now we
can take a look at them. •
[Paul] Our base data model is something we call
CDM. It doesn’t allow blank nodes, and we don’t necessarily
use RDF technology to implement CDM. It isn’t a perfect layering.
The persona data model is not only a vocabulary, it also restricts the
identifiers. The persona model can always differentiate between absolute
and relative. •
[Drummond] We should mark each issue, by category as we
go through them. •
[Paul] I’ve updated the new wiki page during the
discussion. •
[Paul] Global URI’s, if resolvable, resolve to
exactly one description. •
[Drummond] That is technically correct, but multiple
versions may be technical versions for the same thing. •
… •
[Paul] CDM will house multiple values, XDI does
not… •
[Drummond] Subcontext in XDI means a nested graph, not
a named graph….. •
[Paul] I’ve updated the wiki pages as
we’ve talked. We can continue next week. [1] http://www.eclipse.org/higgins/projectplan.php [2] http://wiki.eclipse.org/Higgins_1.1 [3] http://wiki.eclipse.org/Higgins_XDI_Harmonization |