OpenMI Association Technical Committee meeting no 6 Minutes

shopholisticSoftware and s/w Development

Dec 13, 2013 (3 years and 10 months ago)

195 views

OpenMI Association Technical Committee
meeting


1




OpenMI

Assoc
i
a
tion
Technical Committee meeting no 6



Minutes

(version
0
.1
)



Date:



June

18

-

20
, 2007


Venue:



HR Wallingford
,
Wallingford
,
UK


Participants:

Jan Gregersen
, LICTEK

(
gregersen@lictek.dk
)

(
chairman

of the meeting
)


Peter Sinding
, DHI


Water & Environment (
psi@dhi.dk
)

Wiktoria Daniels
, Wallingford software
(
wiktoriad@wallingfordsoftware.com
)

Adrian Harper,
Wallingford Software

(
adrian.harper@wallingfordsoftware.com
)

Stef Hummel, WL | Delft Hydraulics (
stef.hummel@wldelft.nl
)

Onno Rooseschoon
, Alterra (
Onno.Roosenschoon@wur.nl

)


Apologies
:


Peter Gijsbers, WL | Delft Hydraulics (
Peter.Gijsbers@wlde
lft.nl
)





Stefan Westen, Wallingford Software (
stefan.westen@wallingfordsoftware.com
)

Rob Knap
en, Alterra

(
Rob.Knapen@wur.nl

)

Flip Dirksen, RIZA (
p.w.dirksen@riza.rws.minvenw.nl

)


Objective:

OpenMI 1.4

release

OpenMI version 2.0



Plan developments needed for the LIFE project

Discuss cha
nges requests posted on sourceforge



Documents:

http://www.openmi.org/


http://sourceforge.net/projects/openmi


http://www.lictek.dk/openmi/






Agenda /
Issues:

1

Minutes from last meeting

see document:

http://www.lictek.dk/openmi/Meetings/Me
eting05/OpenMIAssocaition.TechnicelComitteeMeeting.5.Minutes02.doc


no comments

2

Use Cases


2
.1

There was no changes to the LIFE use cases since last OATC meeting


3

Feature requests


3.1

Feature requests
at :
http://www.lictek.dk/openmi/OpenMIFeatureRequests.html

and on
http://sourceforge.net/projects/openmi/



All feature requests posted on source forge since last meeting
was targeting the model providers and has been
handled by the model providers.


4

OpenMI 1.4 release


OpenMI Association Technical Committee
meeting


2


4.1

Review the new developers working procedures when using OpenMI 1.4

With the OpenMI 1.4 release developers need to make their own compilation of the
OpenMI SDK and also need to
take care of installation of the associated assemblies. This is a new way of working as compared to previous OpenMI
releases. Before the final OpenMI 1.4 release we must make sure that this way of working is really what we want.


Once the full first version of the OpenMI 1.4 installer is completed
everyone

should try to work with their model
migration in the new SDK way and gain some experience.


In progress. Peter

Sinding
has already tried the new procedure using OpenMI 1.2, and

things seems to work as we
would like them to.
Everyone

should keep doing this until the next meeting.


4.2

TreeView performance in OmiEd (Adrian)


I’ve

been working on performance improvements to our IW implementation of OpenMI for the real life size
m
odels being used in LIFE.

This was an issue as raised by users at
t
he last workshop. I have for CS improved our ExchangeItem
management so now not the critical path.



The problem is now how the OmiEd treeview control performs slowly for many nodes, it is
very poor and
gets worse as, for example, for a component you keep closing and reopening model properties on a
loaded component.



The performance of .NET TreeView has been complained about on the web see

http://www.dotnet247.com/247reference/msgs/11/59000.aspx

for example.



I did a play with changing the TreeView so it only loaded on node expansion and deleted on

collapse.
Improvement measuarable but not significant.



This

can be argued as

a design issue with OmiEd in that TreeView's are not appropriate for large datasets.



As it currently stands OmiEd is not usable for generating data connections for large mod
els, the graphics
response is prohibitive.



With my WSL hat on I can get around this by letting the user filter the selection (by region say) within
InfoWorks and then make the LinkableComponent only publish that subset within OmiEd. Indeed, I will
have t
o do this to get are UserCase's workable quickly.



The above might be
considered

an
expectable

long term solution, but I suspect not. Therefore I suspect
that we will need to revisit the OmiEd design.



I have convinced myself that I can dump thousands of

exchange items into an OmiEd, but the OmiEd must
then not use fancy controls to process them.



PS my tests were with 67000 exchange items, usercase's A and B will require at least 5 times that many.


OATC recognizes this
as
a significant problem that ne
eds to be solved e.g. in the 1.4.1.0 release of the
GUI.

The problem could be solved by changing the design of the link
-
dialog in the GUI, so it uses simple
lists rather that tree views. Selection of elementsets and data operation would then be done using
e.g.
combo boxes.
Jan

will put this issue on the agenda for the September meeting.


4.
3

OpenMI 1.4 release schedule


OATC May meeting (Week 19)


Week 20



Week 21

XML comments completed, and checked into source forge svn.



OpenMI Association Technical Committee
meeting


3


Utilities.Wraper, Utilities.Spa
tial, Utilities.Buffer (
Peter S

in progress
)


Standard (
Jan

in progress
)


Jan

has completed the documentation review and has sent the comment to Peter G.

(
Completed
)

Peter

will upgrade the OpenMI UML diagrams on source forge (logo and version stuff).
(
Com
pleted
)



Week 22

XML comments for the Backbone completed (
Adrian
)

T
he Alpha release is completed and send around. (
Adrian
)



Week 23

Everyone

is testing the alpha release and reporting back to Adrian


The OpenMI Standard definition version draft document

completed and distributed to AOTC. (
Jan
)

Week 24

Adrian

will complete the Beta installer and ask people for things… (e.g. completed docucuments)


NTUA

will complete the download facility


Make the beta version of the installer available for OATC (Wedne
sday). (
Adrian
)



OATC June meeting (week 25)


Final testing. Testing on different operating systems (XP pro, XP home, win 2000, win 2003 server, Vista)


AOTC members upgrade their own models to comply with OpenMI 1.4 and test if thing work as they should
. This
also includes creating installers for these products.


OATC September meeting


Go or no go for the release.


Put information on the OpenMI web about the release


OpenMI 1.4 release uploaded to
www.OpenMI.org




MINUTES FOR 1.4 Release


Installation




Check the LGPL text in the installer

(
Adrian
)




The “required” term in the installation dialog for OmiED should be changed. To something like: “which
support OpenMI linking…”. This is done in order to emphasize that
the GUI is not required (not part of the
standard).

(
Adrian
)




The “required” term should also be changed for the SDK installation dialog. We should use “Supports
developers….”.

(
Adrian
)




Adrian

will try
to
make facilities to decide where each module is ins
talled (Standard, GUI, SDK). By default
everything is installed into programme files.




Check that SDK works with visual studio express.




What about installation of NUNIT
. It should be
recommended

in the Install info file

to

install NNuit
, newest
version,
2.4.1
.
There

should be two build options: one with everything and one without unit test. The one
OpenMI Association Technical Committee
meeting


4


without unit test should be the default one (same as “debug”).

Adrian





Adrian

will remove the icon files.




We need a version of the standard.chm file to be i
nstalled with the standard. The
re

should be a menu item
for this also.

(Adrian)







Adrian

will add the
separate “Standard” item to the start menu. From this menu there will be two sub
menus Help (which redirects to chm for the standard) and Documentation.




Check the GUI documentation (some licence and some links that needs to be reviewed).

(
Jan
)




Check the help about dialog in the GUI. (version number and original authors )

(
Jan
)




Find some example files for the GUI and a
create
HTML file explaining the ex
amples.
(Jan)
Jan will mail
such directory to Adrian. A menu item under that GUI start menu (examples).




Add an assembly folder containing, the org.OpenMI.Standard.dll. Also make a registry key that point the
this folder (as we did for version 1.2).

(
Adria
n
).




The Standard project should be moved from the SDK folder tree to the programme files/ OpenMI 1.4.0.0/
standard/source (
Adrian
)




All references to the standard.dll in the solution file should reference the GAC. All
other
references
should
be reference
s to projects. (
Done

)





We decided that we to not need at merge module for installing the org.openMI.Standard.dll.




Standard XML comments:
Jan

will complete the XML comments

for the Standard
, when done mail to AOTC
to review, and make private notes. At t
he next OATC meeting we will go though it again.
Adrian

will make
sure that the HTML tags are used when making the chm file (such a line breaks).





We need to create a install info file, which describes the new way of model deployment (see use case
below).

Jan

will write this document.




Copyright statements in all source code files should be changed. The ILinkableComponent.cs file has the
new statement; all other files should have the same statement. (
Adrian

will do this)




Move the LinkableComponent XSD fil
e from the standard. Programme files 1.4 standard / bin
(
Adrian
)




Make sure that the OpenMI GUI provides a nice message when non OpenMI complaints models are loaded.
(
Stef
).


How we think the SDK typically will be used (use case):


Primary actor: A develop
er that are creating an OpenMI compliant component.

(PA)

Secondary actor: A user (third party) AU.


1.

PA download OpenMI SDK from
www.OpenMI.org

2.

PA installs OpenMI SDK on her PC

3.

PA Opens the SDK solution

4.

PA Compiles the

debug solution (the default one)

5.

The compilation completes without errors and warnings

6.

PA saves the solution as MyLinkableComponent.sln

7.

PA creates a new project in her new solution

8.

PA adds references to the standard (the GAC) and the utilities
, backbone,
developmentSupport

projects

9.

PA creates a class that implement IEngine and a class which inherits the Wrapper.LinkableEngine.

10.

PA does a lot of programming…….

11.

PA compiles

12.

PA creates a zip file containing the MyLinkableComponent.dll, the org.OpenMI.Backbone.d
ll and all other
OpenMI Association Technical Committee
meeting


5


files needed.

13.

AU receives the zip file, Unzips the file

14.

AU runs the MyLinkableComponent user interface, populates the model and saves. This gui the saves a OMI
file

15.

AU downloads the OpenMI standard from
www.openmi.org

and install the standard.

16.

AU downloads the OpenMI GUI and installs it

17.

AU loads the
OMI

into the GUI

18.



Alternative path:


1.

13a PA also includes the signed OpenMI.Standard.dll





OpenMI 2.0 architecture


5.1

OpenMI version 2.0 roadmap

OATC May Meeting


Iteration 1 implementations: Target is to ensure that the interfaces work as we would like them to. No fancy stuff,

Iteration 1: focus is on the configuration phase


Iteration 1 Architecture (completed)

Iteration 1 implementation of Back
bone, Development support (to be merged with the backbone). (
Stef
)

Iteration 1 implementation of Wrapper, Spatial, Buffer (
Jan
)

Iteration 1 implementation of testmodels (e.g. Simple River). (
Jan
)

Iteration 1 implementation of GUI (
Gena
)


OATC September M
eeting


Iteration 2: The Object as return type from GetValues, Persistent state


OATC October Meeting


Iteration 3: ElementSet,

Events. Draft OpenMI 2 architecture document (draft).


OATC December Meeting


March 2008 OpenMI 2.0 standard interface freeze


September 2008 OpenMI 2.0 beta internal LIFE release.


Minutes:


Peter has upgraded the wrapper component to 2.0. It was possible to keep the IEngine interface as it is. However, for
the final release most probable we will also make a 2.0 dedicated optimi
zed wrapper.



I was decided to add the IElmentSet and the IQuantity to the IOutputItem. Main reason for this is to enable the user
interface to provide better information for the user when she is creating the link.


Peter also
migrated

some of the simple

test models (those in the wrapper unit test folder).


Details on next iteration (Iteration 2) (from now until the September meeting):



The GUI was not implemented in the current iteration.
Gena

will implement the GUI in iteration 2.



We have a nice facilit
y with respect to outputItems, we foresee that a similar mechanism is needed for
inputItems.
Everyone

thinks about this and
Jan

will put this issue on the agenda for next meeting.



We need to elaborate on the returning object thing.

This issue also includes

the discussions about quantities
and qualities.



We need some use cases:

o

Use case describing the quality / categorized issues (
Stef

will mail Rob and ask him to provide a
use case)

OpenMI Association Technical Committee
meeting


6


o

Spreadsheets, flat file, components without time, analytic
models, GIS, ec
onomic models.
(
Everyone
)


o

Usecase for groundwater (MODFLOW) and river and WQ. (
Stef

& Jan
)


6

OpenMI Java and OpenMI .net synchronization


6.1

From Onno:

That we have different opinions seems clear. ;
-
) We'll have to try to solve any problems we have.
Jan
-
Erik had a 30
minute talk with Peter Gijsbers on the phone yesterday about how we work together. Alterra is a new partner in the
OpenMI tech committee and we have a different background when looking at OpenMI then the other partners. And
we see the fut
ure a bit different also so it seems.



Nevertheless, we have to discuss how we can work together to the common goal of spreading the use of OpenMI.
But, when talking about strategy, that the main task of the OAEC. In our opinion the OATC should carry out
the
strategy of the OAEC. What we should do, is discuss how the TC should work, what the responsibilities and roles of
the different members should be, and how we communicate. Communication to both the outside world, eg through
sourceforge an the OpenMI.or
g website, but also between the members of the OATC. I think that how we
communicate lately, is the cause of misunderstandings and

disappoinment and

sometimes even anger. This must be
avoided if we can. As I heard from Jan
-
Erik, Peter supports the need to
put this on the agenda next week.



So I propose to start the meeting from next week with

a discussion on a OATC charter or how you want to call it, in
which we clearly define how the OATC is structured, how and when we communicate with whom, how desciscio
ns
are being made and when they maybe presented as an official OATC statement, representing the opinion of all
members.



But for now

the most important thing is a message on Sourceforge in which a clear statement is made about the still
supported (by OAEC
) use of Java, as reffered to in my message from Monday. This has to be done now, because
current info in incorrect or at least

incomplete and does damage to OpenMI. We truly cannot wait until next OAEC
meeting. So I would like to hear from you when we can

expect an additional news item on Sourceforge on this?


Actions

Lots of discussions but no real agreements.


Jan

will remove the short statement on source forge. The long story should stay, but with different title (
For version
1.*
Only C# is supported
, b
ut for future releases Java implementations may be supported.
) .


On longer term we need an official strategy document that the AOEC fully supports. The starting point for this
document should be the strategy paper by PG

and the long story on source forge

with additions made by
Onno
. On
the September

OAE
C meeting we should decide on the strategy document.


We need formal guidelines for how to communicate on source forge and on
www.openmi.org

E.g. how to make
statement
s on behalf of AOEC and OATC. This is also an issue that should be handled by AOEC and AOTC.
Jan

will
ask Roger to put this issue on the agenda for the next
OAEC

meeting and
Jan

will put this issue on the AOTC agenda
also.


Onno had a meeting with Roger ab
out these issues. The decision was that
Onno

will make a document based on
Peter G. s document and circulate this document to Roger, Peter G, and Jan.


Other JAVA /. Net issues.


On short term w
e should investigate how to make a Java equivalent of IEngine.

(
Stef
)

On longer term (which relates to OpenMI 2.0, we should investigate a full bridging approach. Gena is already doing
some investigations about this.


7

Miscellaneous issues


7.1

Subversion issues (Gena)


Please see some ideas about repository and
release planning. It is generally how subversion is used by many OSS
projects.

OpenMI Association Technical Committee
meeting


7



... another comment is about idea to put SDK into "My Documents" ... probably would be better to keep it together
with other folders:

(Program Files/OpenMI/).

As an example che
ck Where .NET or any other SDK is installed.


The most important from my comments is that we should have
one folder containing main development line
and
folders which will contain released version. This approach is used by most OSS projects. There is also
recommended
repository layout for Subversion which uses
branches/, tags/, trunk/
folders. Usually almost all Subversion projects
are defined using these folders.

The real project tree stays under trunk/ folder and it is current development version of sourc
es. Tags and branches are
used to store, well, tagged and branched versions of trunk.


OpenMI/



branches/



tags/



release
-
1.0/





src/







doc/







...



trunk/
-

here main development occurs, when new version is released
-

it is copied into tags/ folder using
command (1)





src/





doc/





...


(1) svn copy trunk tags/release
-
1.0


It should be considered as conventional agreements, but it is good to use them so that the layout will be clear to
anyone who is fam
iliar with open
-
source projects.


For example check these:


http://anonsvn.jboss.org/repos/hibernate/

-

very popular ORM library.

http://fisheye.codehaus.org/browse/jruby

http
://fisheye.codehaus.org/browse/xfire

-

note that there is the same folder in branches/ and tags/ called xfire
-
1.1/. It
is because files in tags/ should never be changed and when changes are required
-

it is copies from tags/xfire
-
1.1 to
branches/xfire
-
1.1

and changed there. The log also shows that version in branches/ was changed 11 month ago and in
tags/
-

12 months.


Please check these links for details:


http://svnbook.red
-
bean.com/nightly/en/svn
-
book.html#svn.branchmerge

http://software.ericsink.com/scm/scm_branches.html


OpenMI Association Technical Committee
meeting


8






We like the idea of

working with svn in the “real

way

. However, since we are very close to the release
Adrian

will
do as follows:


When Adrian is happy about the files for releas
e 1.4 he will check these files in as the main OpenMI trunk and tack
this as OpenMI 1.4Beta. From then on we will work in the “real” way.

OpenMI Association Technical Committee
meeting


9



We discussed changing the folder structure so it will reflect the differences between SDK and GUI. However, we
want
it to be easy for people to upgrade to 1.4, so changing namespaces is not an option. We like the namespaces to
reflect (be the same as) the folder names, therefore we cannot change this either. So we keep tings as they are.


8

Matters arising from last mi
nutes of meetings


8
.
1

From OATC meeting 4 m.o.m. point 4.

The performance tests are done using models that all are running on a single PC and in the same process. We
envisage that situations where models are communicating between computers or between pro
cesses will have a
higher performance penalty. We realize that cross process communication will have a performance penalty no matter
what technology is being used (OpenMI or other protocols).
Stef

will make further investigations of this


Ongoing


8
.
2

Fr
om OATC meeting 4 m.o.m. point 4.

Jan

and
Peter S

will fine
-
tune the performance test document and make it available on the
www.openmi.org

web
site.
Jan

and
Peter S

will also consider turning the document into a confe
rence paper.


Ongoing



8.
5

From OATC meeting 4 m.o.m.
point 6.1

Stef

will investigate JN
-
Bridge
www.JNBridge.com

for interfacing.


Stef

will together with Gena continue this work


Ongoing (but remove from list si
nce this is also described elsewhere)


8.
6

From OATC meeting 4 m.o.m.
point 7.1.3

Adrian

will make an installer for OpenMI 1.4


Done


8.
7

From OATC meeting 4 m.o.m. point 7.1.3

Jan

will write a OpenMI standard definition document


Ongoing


8.
9

From OA
TC meeting 4 m.o.m. point 7.1.5

OMIEd should, by default, run each component in a separate process.
Adrian

will look at the implementation of the
GUI in order to investigate how such facility can be implemented


Ongoing. This feature will not be included i
n 1.4.0.0 but in 1.4.1.?


8.1
0

From OATC meeting 4 m.o.m. point 7.1.5

Jan

will add to the standard requirements that OpenMI compliant components must be able to run as multiple
instances in the same thread.


Should be both i
n the xml comments and the Open
MI standard

definition document


Ongoing


8.1
1

From OATC meeting 4 m.o.m. point 7.1.5

We

should make some guidelines for implementation of LinkableComponents that are running the engine in a
separate thread. Such guidelines should be published on
www.openmi.org



Ongoing


8.1
3

From OATC meeting 4 m.o.m. point 8.6

OpenMI Association Technical Committee
meeting


10


Backbone.Quantity constructor should use IUnit Currently the constructor of the Quantity
-
object
(org.OpenMI.Backbone) uses at the same time a Unit
-
clas and
an IDimension
-
interface. The constructor should adopt
the interfaces (and thus IUnit) instead of the classes. Implementation is done but
Peter S

will make sure that all Unit
tests run without any errors.


Ongoing


8.1
4

From OATC meeting 4 m.o.m. point 8.8

From Wiki: If I have RS running against CS, and RS throws an exception (say the solution becomes unstable), then I
am able to shut down the RS model. However, then I am still left with the CS model which sits there waiting for
something to happen. It wo
uld be nice if I could send some sort of event which would then be caught by the GUI e.g.
the simulation failed event, and then the GUI could handle things appropriately. This would involve adding the
simulation failed event to the list of standard events

(a possible feature request). This would all be very well, but all
my exceptions are thrown from a class which implements IEngine, and I can't work out how to send events from
IEngine
-

is there any way of doing this? (see
minutes of technical committee
meeting no

8.3)


Wiki

and Adrian
will

make some further investigations of this.


To be removed from this list.



8.1
5

From OATC meeting 4 m.o.m. point 8.
10

Flip has mailed the datamonitor code to Wiki. Wiki will test the code and when thing are working
A
drian
will
upload the code to sourceforge OpenMI version 1.3.


We decided not to include the data monitor in version 1.4.0.0. We think that it needs to be improved before it is
release. We are considering releasing it with 1.4.1.0. Remove this item from th
e list.


8.1
6

From OATC meeting 4 m.o.m. point 8.
11

Jan

will make sure that the data monitor is included in the installation for the next release.

(see change request 12)

The data monitor should be included in the GUI installation.

Adrian

will take care
of this when he

is creating the installer

To be removed (see point 8.15 above)

8.1
7

From OATC meeting 4 m.o.m. point 8.
14

Rob

will provide an example for pre
-
initialize and providing arguments to Linkable Components

Ongoing


8.
18

From OATC meeting 4 m.o.
m. point 8.1
5

Wiki and Adrian

will investigate how the existing wrapper can be modified to support parallel

(See minutes from meeting no 3 point 5.6)

Ongoing

8.
19

From OATC meeting 4 m.o.m. point 9.1


Jan

will investigate how OpenMI compliance test tool c
an be made.


Ongoing


8.2
0

From OATC meeting 4 m.o.m. point 9.
2

Design pattern for remoting. Both design pattern and implementation in the wrapper (cutting at the IEngine level) and
also describe alternative ways of cutting.
Adrian and Stef


Ongoing


8.2
0

From OATC meeting 4 m.o.m. point 9.
2

OpenMI Association Technical Committee
meeting


11


Ask David Ma
idment about how they made the web service (
Jan
). Design pattern and implementation


Ongoing


8.2
1

From OATC meeting 4 m.o.m. point 9.
2

Design patter for
Wrapping stateless models. (
Peter S
).


Ongoing


8
.22

From OATC meeting 4 m.o.m. point 9.
2

Running OpenMI on Linux (e.g. using mono). (
Peter G
. will ask Gena about this (maybe to present something on the
next meeting)


At the next meeting in UK the focus should be on the 1.4 release. We try to get this
pr
esentation at the OATC
October meeting… (
Peter G
.

will

inform Gena about this).


Ongoing


8.23

Actions related to OpenMI 1.4 release (from m.o.m meeting 5).




Update XML code comments

We should make sure that the XML code comments are updated and a documen
t which provides easyaccess
the these commets is generated and installed along with the OpenMI 1.4 release. We may use “NDoc” to
g敮敲慴攠th楳 do捵men琮tt攠w楬氠i楳瑲楢u瑥tth攠upd慴ang work among us as fo汬lws. (no瑩捥 瑨慴aw攠
d敬楢敲慴敬a d散楤敤 瑯 sk楰
som攠p慣kag敳e


啴楬楴楥s.tr慰敲I 啴楬楴楥献印慴楡氬 啴楬楴楥s.Buff敲 (
Peter S.
)

Backbone (
Adrian
)

Standard (
Jan
)

DevelopmentSupport (
Jan
)




Stef

will take care that equals and the get hash code are update for the Backbone.




Remove the command line menu i
tem from the installer. (
Adrian
)




Spilt the HTML documentation main page into SDK and GUI / Standard. (
Adrian
)




Test if a merge module is needed for the standard or if software provider can include the standard.dll into
their own install script. (
Adrian
)




Check if assembly folder is needed in order to allow developers to reference the Standard.dll (
Adrian
)




Change the logo in the installer (the LIFE OpenMI logo should be just the OpenMI logo) (
Adrian)




Remove the standard dll from the GUI bin directory. (
Ad
rian
)




Adrian

will get complete beta version of the installer in about two weeks after the meetings.




Adrian

will write at section to the management protocol document explaining the procedure for creating an
installation.


Some of this has been done and s
ome not, see new action list for release 1.4 under point 4.3 above


8.24

The new OpenMI standard definition

The OpenMI 1.4 release comes with a more strict definition of what OpenMI compliance means. Jan will make the
first draft of this document. We mus
t make sure that the overall content of this document is ok.


Jan

has started writing the xml comments and will complete this before the next meeting. When done
Jan

will e
-
mail
the OATC and ask for comments. We will discuss this also at the next OATC meeti
ng.


OpenMI Association Technical Committee
meeting


12


Ongoing


8.25

Documentation

Because the OpenMI 1.4 release imposes new working procedures for developers we need to review the current
documentation and possibly make some changes. Alternatively, we could describe the changes procedures in a
release

news document.


Jan

will continue the review of the documentation and mark the sections where changes should be made.
Jan and
Peter G.

will subsequently find out how to distribute the task of updating these documents.


Done


8.26

OpenMI installation do
wnload from
www.openmi.org


At the last OATC meeting we decided to move the OpenMI installation download facility from source forge to
www.openmi.org
. This should enable us to mo
re closely to monitor who is downloading OpenMI.


Jan

will write a mail to Ria explaining the details (e.g. only registered user are allow to download, the web should
keep a record of downloads) and
Jan

will also emphasise that this facility must be ready

for testing before the June
OATC meeting.


NTUA is working on it and John has suggested at solution.
Jan

will mail back to John and ask him to complete the
download facility the way he suggests.
As a start we will make the OpenMI 1.2 available through the

new download
facility. When 1.4 is ready this release will replace 1.2.




Matters arising from the OpenMI LIFE use cases and the OpenMI LIFE workshop


We discussed the conceptual aspects of the Scheldt Use Case C? Infoworks RS to PEGASE.
Jan

will write

at
document with suggestions about the linking can be handled by use of mapping rather than using ID
-
based links.
Jan

will send this to Yves and to OATC.


Jan

has a task from the LIFE SC meeting to contact Costas and ask for some information about his wor
k


Ongong


9

Any other business


9
.1



9
.2


9
.3




Comment to the minutes of meeting by Rob Knapen:

Dear Jan,



Here are

a few comments on the minutes. Since I could not attent the meeting they are more general comments
than

being about

wether the mi
nutes reflect what was

discussed or not. Hope they are still useful (although they
almost all seem to bring up the .NET / Java discussion, sorry for that). I have numbered them according to the
agenda items / issues:



2.1. Use cases compared with the minu
tes from the previous meeting seem now to be limited to the LIFE use
cases. Is this for the 1.4 release only, or also for the 2.0 release?



3.1. Feature requests: The list OpenMIFeatureRequests.html was last updated on January 31, 2007 and for
some of the

requests there is still no clear answer. Could this be fixed?



4.1. OpenMI 1.4 release: Since Java is not included I can not test it with our current models. Also it is not
possible to migrate

these


models

(unless we would rewrite everything

in .NET, wh
ich would be too expensive to
do).



OpenMI Association Technical Committee
meeting


13


5.1. OpenMI 2.0 architecture: At the moment everything is done using C# and Visual Studio, so it is difficult to
consider the impact of going from the Java 1.2 version to a compatible 2.0 version. It would be good if th
e design
documents (UML's) would be available in a generic format (e.g. PDF). I also think that the roadmap should
clearly show milestones for the development of the Java version


(needs input from Alterra).



5.1. Quality / categorized issues: A sample us
e case would be a simulation model that predicts land use over
time, so for an element set of spatial grid cells for each cell it returns if it will be "grass", "built
-
up area", "water",
"agriculture", "infrastructure", etc. These are qualities, without lo
gical order. Or or model that outputs the taste of
food ("sweet", "salty", etc.) given a number of ingredients.



6.1. From a practical point of view: I (and other people) are using and need the Java version of OpenMI for
projects. .NET and Mono is no usab
le alternative at this point. There is also a need to continued development of
the features of this Java version, because of requirements of the projects and deadlines. And finally some of
these EU funded projects strongly support (or require) that source
code is made available as open source
(including the Java OpenMI code). If all this can fit into the OATC and OAEC strategy, so much the better (for
all)...



7.1. Subversion issues: In general it is a good proposal, but there is no room to store source co
de in multiple
programming languages.



8.17. An example for pre
-
initialize with arguments: Suppose my linkable component uses init arguments to load
a general scripting engine (or formula processor or something like that) which in turn needs to be initial
ised with
arguments (e.g. to load a script from a given file). The second set of (valid) initialisation arguments is only known
after the first initialisation is done. Also this might be further nested, requiring some kind of iterative initialisation,
like
:





repeat



{




args =

linkableComponent.getInitArgs();




... fill in the args ...



} until (linkableComponent.init(args) == INIT_COMPLETE);





8.24. The new OpenMI standard definition: Personally I am not in favor of making the stan
dard language
(implementation) specific. However, looking forward, how will this 1.4 version be updated / modified so that the
2.0 version includes both a .NET and a Java implementation?






I was hoping to be able to attend the next meeting, however it l
ooks like it is in the middle of my vacation (which
is early September).



Kind regards,

Rob