Attendees
- Brian Carroll - Serena
- Andy
Hodgkinson - Novell
- Drummond
Reed -
Cordance
- Mary
Ruddy
- Meristic/SocialPhysics
- Markus Sabedello - Parity
- Jim Sermersheim - Novell
- George Stanchev - Serena
- Paul
Trevithick -
Parity/SocialPhysics
- Brian
Walker -
Parity
- Hank Mauldin - Cisco
Time: noon EST
Dial-in:
1-866-362-7064 / 892048#
Agenda
1. [Brian] 1.1M4
was made available on November 25th
- http://wiki.eclipse.org/Higgins_1.1M4
- http://wiki.eclipse.org/Higgins_1.1M5
– now working on this milestone
- [Brian] Valerie sent a
message last week about the M4 update. M5 is tentatively set for January 9, a
little more than 6 weeks. There
is an extra week due to the holidays. Moved all the unfinished M4 items to
M5. On M5 list, selected some of
the higher profile items that are actively being worked on for M5.
2. [Paul] Java 1.4 vs..
Java 5
- We’ve taken a vote/poll
to get a sense of the group.
- Wayne Beaton emailed Mary
and I privately and said that he felt that the +1/+2 thing was confusing and
recommended that it should have been done as two separate votes. He’s no doubt
correct. My apologies for not creating two separate polls
- Jim also said the same
thing: he felt there was no way to convey that he preferred +2, but if that
fails +1. I think most everyone felt that way
- Results:
- Those for moving
everything to Java 5 (+2): Markus, SergeyL, Jeesmon, Daniel, Tom, Paul,
Maxim
- Those for allowing
selected components to use Java 5 (+1): <same list as above is at least
partially implied>
- Those for not allowing
any use of Java 5 (-1): Paula, Mike, Greg
- Paul & Mary: We’re
leaning towards +1: “allowing selected components to use Java 5” because
unlike many other decisions we make, this one may mean that an entire set of
developers (e.g. IBM’s developers and
potentially others) are disenfranchised. They will find it harder to justify
contributing to the project as they believe that it will reduce Higgins 1.1’s
adoption due to incompatibility with the installed based of 1.4 compatible
JVMs. And as we all know Higgins needs all the volunteers/resources it can
get.
- [Paul] I put a lot in the notes for the agenda
that you can read. There were some structural flaws with the way I created the
vote and I apologize for that. The last paragraph is where Mary and I settled
out. If we vote +2 it really makes it difficult for some to put resources into
Higgins and that is a very serious down side
for us. On the other hand, there
are some components that would really benefit from Java 5.
- [Hank] How will that really work? Which components would be
allowed? New optional
components?
- [Paul] For new optional ones, that’s easy to do
with Java 5. For core items, 1) there is always the option of using
Higgins 1.0, which runs on 1.4 or 2) would
need to use retrofitting tools or not run it. I hear what you are saying. Means
can’t say that it runs on 1.4 if all of it doesn’t. The other approach is to
support both. A component owner could use 1.5, but would still need to support
1.4.
- [Hank] I’m thinking how easy this would be if we
allow some to move to 1.5, while others stay on 1.4. How would they build it? If only support one version of Java,
than the users can’t move forward.
Maybe that is part of IBM’s point,
if you only offer one version, of course the customer doesn’t move. I’m not
opposed to moving forward, but am concerned about some halfway points that may
make it more difficult for customers.
Move forward, or support two if need to.
- [Paul] There might be a distinction between
components that run on the server, vs. the client. Might want to be more
conservative on the client.
- [Jim] I think the mixed version might work. Keep
some core components at 1.4, while allowing some other things to be a 5, but
the hard part is we don’t really have a dependency graph. You can’t move the
configuration code or IdAS, things that are depended on by other projects. It
seems it would necessitate us to move to a make it clearer what the dependency
graph is.
- [Hank] Is there enough backward compatibility
that anything running in 1.4 would run in 1.5?
- [Jim] That is my understanding.
- [Hank] Just wanted to make sure.
- [Paul] That’s right.
- {Paul] We need to define more clearly what
allowing selected means, and start making core vs. not core clear, and make
dependencies clear. Break the
dependence project and break it into individual projects.
- [Jim] If we do the +1 thing, it does mean that
all of the benefits that Markus enumerated, we wouldn’t be able to do any of
that in the core modules. Which is probably fine give how much time we see
people putting into refactoring. There are some things that would be very nice
– like generics so have type safety.
- [Paul] This is also a decision, where we are
taking the smallest possible step forward. We can decide in the future to do
something different for 1.5. This
is a tough one and passions run high. We have to feel our way forward.
- [Hank] Maybe an action it should be to
test that [backwards compatibility].
- [Paul] I think that will effectively
happen.
3. [Mary] Website home
page progress update.
- http://www.eclipse.org/higgins/nursery
- Partially complete new
home page; still broken links, etc.
- [Mary] We have the new graphics for the new home
page based on feedback from the F2F and IIW.
- [Hank] Love them. Seems great.
- [Paul] Think there is
something wrong with images. I
seem them render slowly
- [Paul] From before they
were very huge. Thought they
would be scaled down.
- [Mary] The new images are smaller than the
first versions.
- [Paul]- Maybe it is my
Internet connection.
- [Drummond] They are slow
for me too.
- [Markus] The size is very
high resolution.
- [Paul] See, they are
still huge. Need to be
resized.
- [Paul] I also think looks
great. Concept is to highlight
IdAS and bring it out separately.
We also discussed that door number 1 may not just be i-card
selectors. It may be also include
the selector selector. Or there
was some interest at IIW – growing interest - on converging on standard
plug-ins. We need to highlight
that also.
- [Mary] The next step is
to fix the text and links.
- [Paul] I also like that there is an overview
page.
- [Drummond]
Yes.
- [Paul] Any other
comments? Jim, do you like it?
[Jim] Yes. It’s just that the background is curved
on the first picture and not on the new pictures.
- [Mary} I will get that
fixed.
- [Paul] I find that the
grey text is hard to read. It
needs to be a little darker.
- [Drummond] I
agree.
[Mary] I will fix that.
- [Paul] Great progress.
4. [Brian]
STS PPID algorithm
update
- SergeyL has completed and
tested a patch to org.eclipse.higgins.sts.client
- It is compatible with
Microsoft’s CardSpace implementation
- Sergey has requested that
Mike review the patch and commit it if he has no
objections
- [Brian] Asked for anyone
else to take a look at it before committing the patch. If you note on the M5
wiki pages there is also another bug fixed. Will include that when get patched
reviewed. We do need some review.
On our side it tested out
ok. So provide feedback or
questions to the dev list.
5. [Paul] Higgins
Selector Architecture Harmonization
- We’ve learned a great
deal about building selectors by rapidly building 4 different flavors
- We need to converge on a
single architecture
- Need to define common
component APIs (service descriptions AND component names!)
- Some of us in Parity,
Novell, as well as potentially some independent developers and other
organizations, are interested in starting work on this.
- We’d start by unifying
GTK/Cocoa C++ Selector with
the AIR Selector
- http://wiki.eclipse.org/Selector_Architecture_Harmonization
- [Paul] The selectors use a lot of different
components. So if we could harmonize the selectors, it would go a long way to
harmonizing all of Higgins.
- [Paul] Move towards common components and maybe
have multiple implements of each (java, and C++).
- [Paul] As noted, there are many people at Novell
and Parity, etc. who are very interested in a converged architecture. Could as
a first move make a GTK and
AIR version. If you click on the link, you see the
new diagram….
- [Paul]The concept of a package is I think a
missing ingredient in our architecture.
A package supports a service, but that package could be implemented by
different components. Say a package of selector selector – on Windows it is
implemented by one set of components; on a Mac it is implemented by another
set of components. Sometimes one
component represents one package. Once I introduced that concept, it made
things much clearer. We kind of had this notion, we had bunches of components
on the components page. The
selector UI – DigitalMe, and Parity uses
AIR, logically it is all selector UI. There may even be alternative
components on a single platform. On Linux for example. You can use this
package concept to talk about a Higgins
Identity Server. I haven’t gone back and refactored all the wiki pages. In
addition to all the issues, we have an architecture that is so complex that it
doesn’t fit on one page. With this approach, can have a diagram and then drill
down. May have a package within a package. I dream of a day when we can have a
clickable architecture picture and navigate to different implementations on
different platforms. We used to have a unified architecture, but as we
branched off to quickly implement, we couldn’t do this any more. I’m feeling pretty good about this
approach to managing complexity. Do people have a better word for
package?
- [Drummond] I like package. It is a UML word
also.
- [Andy] There is a green line from selector
client to the identity service. That doesn’t imply does it that the client may
have some smarts to talk to a remote
STS.
Is that right?
- [Paul] Yes, one of the benefits of this package
approach is you can hide a lot of stuff. I you were taking this approach, you
would just call the remote STS your
self.
- [Hank] Is everything in the blue one package? or
is each box one package? I’m working with a media group and think of package
as bundled together.
- [Paul] That is not what I meant. A package is
bigger than a component. Either multiple components or multiple implementation
options.
- [Hank] My only objection is that package is used
in other ways in the projects I work on.
I don’t have a better word now. I will think about it.
- [Paul] The architecture is one where there is a
fairly smart selector client doing quite a lot on the device. We got
encouraged when we hear that someone had got this working on an
iPhone.
- [Paul] There is also a missing package. There is
also an i-card manager web-app, that should be sitting with a green line down
to the identity server also. That is a detail. So Andy, are you otherwise ok
with this basic approach?
- [Andy] I think so. We may want to show a line
going from the Higgins identity server over
to the remote STS.
- [Paul] But the
AIR folks want to move more closely to your
architecture.
- [Paul] We, [the
AIR folks] are looking to use your code.
- [Paul] When I tried to draw the selector client,
I found that there is more plugability in the Java client. Your selector
client was more purpose built. What I would hope is we could evolve to a
fairly modular common architecture, even if in many cases on the client the
plug-in point is never used or thin.
- [Paul]At 30K feet this is a common architectural
layering.
- [Andy] Ok, I see where you are coming
from.
- [Paul] Mary and I were having a conversion
with a French Consortium, FC2, that is related. They continue to
look at Higgins and have some interesting
challenges integrating with ID-WSF. When we told them about the new hybrid
selector, they said that it lines up with everything we have learned about
what one needs to do in a selector.
Orange has been working on
this. The idea to link to an image of a card – they independently came up with
it. It was good feedback. [That another group independently came up with the
same approach.] It may ultimately mean that there is more development work
from members of the consortium, if all goes well.
- [Andy] So they did this
independently.
- [Paul] Yes. They may want to
collaborate.
- [Paul] Any other topics or comments? See you next Thursday.
- end 12:44