-Anish
--
On 11/24/2010 3:20 AM, Mike Edwards wrote:
OFF3599573.069C076E-ON802577E5.003B5E3B-802577E5.003DE269@uk.ibm.com" type="cite">
Anish,
I learned some more - but that leads to some more comments ;-)
Comments inline as <mje>...</mje>
Yours, Mike
Dr Mike Edwards Mail Point 146, Hursley Park STSM Winchester, Hants SO21 2JN SCA & Services Standards United Kingdom Co-Chair OASIS SCA Assembly TC
IBM Software Group
Phone: +44-1962 818014
Mobile: +44-7802-467431 (274097)
e-mail: mike_edwards@uk.ibm.com
From: Anish Karmarkar <Anish.Karmarkar@oracle.com> To: sca-assembly@lists.oasis-open.org Date: 23/11/2010 07:06 Subject: Re: [sca-assembly] Issue 238 proposal v2
Mike,
Thanks for your comments. My comments inlined below.
-Anish
--
On 11/19/2010 1:59 AM, Mike Edwards wrote:
>
> Anish,
>
> Thanks for working on this.
>
> Comments inline as *<mje>..</mje>*
>
> Yours, Mike
> ------------------------------------------------------------------------
> Dr Mike Edwards Mail Point 146, Hursley Park
> Technical Strategist Winchester, Hants SO21 2JN
> SCA & Services Standards United Kingdom
> Co-Chair OASIS SCA Assembly TC
> IBM Software Group
> Phone: +44-1962 818014
> Mobile: +44-7802-467431 (274097)
> e-mail: mike_edwards@uk.ibm.com
>
>
>
>
>
> From: Anish Karmarkar <Anish.Karmarkar@oracle.com>
> To: OASIS Assembly <sca-assembly@lists.oasis-open.org>
> Date: 18/11/2010 23:51
> Subject: [sca-assembly] Issue 238 proposal v2
>
>
> ------------------------------------------------------------------------
>
>
>
> I have uploaded v2 of the 238 proposal at:
>
> http://www.oasis-open.org/apps/org/workgroup/sca-assembly/download.php/40315/sca-assembly-1.2-spec-wd01-issue238v2.pdf
> and
> http://www.oasis-open.org/apps/org/workgroup/sca-assembly/download.php/40314/sca-assembly-1.2-spec-wd01-issue238v2.doc
>
> If you don't want to be confused with spurious diffs from crossref
> updates please look at the PDF and not the Word doc. The PDF version
> contains pages only from sections 8, 9 and Appendix A. These three
> sections are the only ones that are relevant to the proposal.
>
>
> If you are curious as to where the other spurious changes in v1 came from:
> All of them are from cross-reference updates. 4 of those occurred
> because they used the wrong cross-reference (a cut-and-paste error that
> is also present in WD01). I fixed those 4 cases. The other cases, I
> tried to fix them, but Word keeps updating the auto-generated
> cross-reference text (with identical text) even after I accept the new
> changes -- I can't figure out why Word keeps doing that. I tried using
> OOo, but that made the problem even worse. The diffs showing up in the
> example Snippet references weren't worth fixing, since 1.2-wd01 is a
> branch of 1.1-cd04 and at some point we'll have to migrate to 1.1-cd06
> (or later).
>
> *<mje>*
> *One thing to consider seriously for the 1.2 version of the specs is
> whether to do a*
> *complete migration to OpenOffice from Word format. I note that most of
> the newer*
> *spec documents are in OO format.*
>
> *The time to make such a significant change is at the point where we
> rebase 1.2 on*
> *the final 1.1 spec.*
>
> *I'd appreciate knowing everyone's view on this. I personally find OO
> easier to handle*
> *today than Word, but I realize that may not be the general view.*
> *</mje>*
>
> In any case, till we migrate to the latest version of 1.2, the best
> solution for now I think is to look at sections 8/9 and Appendix A as
> those are the only sections relevant to this proposal.
>
> *+1*
>
>
> Here is the explanation for the proposal (the actual text changes are
> small, but below is the rationale for those changes):
>
> There are three syntactic uses of eventType (or some various of it):
> (a) in the EDL, (b) in the event instance/message, (c) in the filter
>
> (a) The EDL defines a QName for the event type and points to the
> associated XSD. If you are not using XSD, there will some other way to
> define the event type (the example that has been discussed in the past
> is a fully qualified Java class name). This is the EDL eventtype. In the
> proposal it is called sca:eventTypeXSD. If we use the EDL defined by W3C
> WS-RA then we won't be doing any work wrt this.
>
> *<mje>*
> *One question I ask is what the spec would look like if we use the WS-RA
> EDL*
>
Appendix A.13 will go away and our filter eventType.sca would refer to
the WS-RA EDL. Other than that nothing else should change.
<mje>
I expected the filter to simply refer to an event type QName - would this really change if
we use WS-RA EDL ?
</mje>
<ask> No, this would stay the same </ask>
OFF3599573.069C076E-ON802577E5.003B5E3B-802577E5.003DE269@uk.ibm.com" type="cite">
> *One thing in common with our current spec is that it is not clear to me
> where the *
> *EDL XML is placed within Contributions. The current spec is silent on
> this - and that*
> *is not good. I'd like to fix that as part of the resolution of this
> issue.*
Note that I have filed a comment with the WS-RA WG in W3C asking them to
do two things:
a) Specify a recommended/best practice/whatever for the file extension.
Something like .evd. If that happens then it could be treated the same
way WSDL is treated wrt including it in a contribution and resolving it.
<mje>
OK, what I learned from this comment is that you expect the event definitions to be
present in some new separate file, ideally with a fixed filename extension - something
similar to a WSDL.
My idea had been to include the same declarational material in definitions.xml files.
I am OK with the idea of using simple SCA resolution mechanisms to get the
declarations loaded
</mje>
<ask> Yes, that is how I see things working together. Apologies, if I didn't make it clear in my earlier comments. The EDLs should be structured to be independent of SCA artifacts. Especially if we start using the EDL defined by WS-RA </ask>
OFF3599573.069C076E-ON802577E5.003B5E3B-802577E5.003DE269@uk.ibm.com" type="cite"> b) Consider something equivalent to schemaLocation/wsdlLocation.
Currently, it requires WS-MEX to get the EDL.
<mje>
The mention of a schemaLocation style of mechanism has me reaching for the garlic
and a nice big stake - basically no way do I want to see another one of these being
invented.
As for WS-MEX, that is completely unacceptable as far as structuring an SCA application...
</mje>
<ask> I don't want to see WS-MEX either. Although I'm told that WS-MEX has ways to make things simple, but I don't know if that would still require conformance to all of WS-MEX.
WRT, killing the vampire (or Zoombies, as Eric points out): I would like to understand the pain point wrt schemaLocation/wsdlLocation
</ask>
OFF3599573.069C076E-ON802577E5.003B5E3B-802577E5.003DE269@uk.ibm.com" type="cite">
>
> *My proposal is that, as with Intent definitions, it should be possible
> for them to be*
> *declared within a definitions.xml document. This DOES NOT rule out the
> potential*
> *for them to be declared elsewhere (such as a repository), but it since
> repositories*
> *are out of scope of the spec, I think it is necessary to have one
> normative way of*
> *declaring event types, even if products provide other more
> sophisticated means.*
I'm confused as to why this would be in the the definitions.xml. The EDL
files that declare event types (similar to WSDLs that declare portTypes)
could be located anywhere in the contribution or in a different
contribution (subject to our import/export rules). In any case, since
you have already filed an issue for this, we can discuss this during
that issue resolution.
<mje>
I get the idea - it just that it is yet another file type that we must track...
</mje>
> *</mje>*
>
> *<mje>*
> *My second problem with the current EDL is that it is NOT extensible.
> Look at the*
> *EventDefinitions declaration:*
>
> * <complexType name="EventDefinitions" >*
> * <complexContent>*
> * <extension base="sca:CommonExtensionBase">*
> * <sequence>*
> * <element ref="sca:eventTypeXSD" minOccurs="0" maxOccurs="unbounded" />*
> * <any namespace="##other" processContents="lax" minOccurs="0"
> maxOccurs="unbounded"/>*
> * </sequence>*
> * <attribute name="targetNamespace" type="anyURI" use="required" />*
> * <anyAttribute namespace="##other" processContents="lax" />*
> * </extension>*
> * </complexContent>*
> * </complexType>*
>
> *which points at:*
>
> * <element name="eventTypeXSD" type="sca:EventTypeXSD" />*
> * <complexType name="EventTypeXSD" >*
>
>
> *I note that event types can thus ONLY be declared as sca:eventTypeXSD.
> If (say) we wanted to declare*
> *an eventType in terms of some Java Class for example, then there is no
> way to do that. *
>
Do you want the fully qualified Java class to represent the event type
or just the (mapping of) XML schema?
I had assumed (perhaps incorrectly) that folks who want to use Java
classes for events would not bother to use this at all.
In any case, it is extensible because eventDefinition has <any> in it
and eventTypeXSD has minOccurs="0".
<mje>
I take the view that this is like using a Java interface rather than a WSDL.
If you build your stuff in Java, why would you be forced to create an XSD mapping?
So for me there should be an eventTypeJava, which declares events in terms of Java
artifacts. Those SHOULD map to XML/XSD, but the developer should not be forced to
create the mapping manually.
Current extensibility is not suitable for this concept.
</mje>
<ask>
This obviously is a much bigger issue and certainly a separate one. But something that needs to be discussed. This is how I envisioned things would work:
For the assembly part, what is important is the filter -- that is the only place where event type shows up (at design time).
We have a filter <eventType.sca> -- this expresses things in terms of QNames and is associated with XSD-based EDLs (hopefully WS-RA's EDL). If there is a need to use Java-based event types, then I imagine that the filter would be something like <eventType.java> which points to one or more Java classes. There really won't be a need for a separate EDL (similar to WS-RA's EDL).
If we wanted to treat this similar to the way we treat WSDL and Java interfaces (mapability requirement), we can define how the mapping takes place (or rather the spec for eventType.java) would specify that.
IOW, The EDL (<eventDefinitions>/<eventTypeXSD>) won't be used with Java-based systems, but would be mappable to one.
</ask>
OFF3599573.069C076E-ON802577E5.003B5E3B-802577E5.003DE269@uk.ibm.com" type="cite"> > *To do that would require defining eventTypes in terms of some base
> (abstract) type which is the head of*
> *a substitution group, similar to the way we've done extensibility
> elsewhere.*
>
I suppose we could do that, but I'll like to understand the usecases
better. Wouldn't you just use JAXB to do the mapping?
<mje>
JAXB may be used to map at runtime, but it ain't the way the developer would work
</mje>
> *If we /don't /do extensibility then we might just as well call this
> type "eventType"...*
>
We can't, as that is used by the filter. We'll have to come up with
something different.
> *</mje>*
>
> b) the event instance *may* contain metadata for event type (such as a
> header). For example, there may be a eventType header. Note the spec is
> silent on this.
>
> *<mje>*
> *I'm not sure that we can live with this if we start to concern
> ourselves with interop.*
>
> *For example, if we happen to send events over either Web services or
> via JMS,*
> *can the spec be silent about a) whether "EventType" is transmitted and
> b) where*
> *it is transmitted? My view is that we cannot.*
> *</mje>*
>
In the past we have agreed to stay away from defining standard metadata
(or any serialization issues). I suppose we can change our minds and try
and tackle this, although I am very wary of this. But in any case, I
think this is a separate issue.
<mje>
It may be a separate issue, but if we are going to concern ourselves with interop - and
I think that we should - then we can't avoid discussing the issue of how metadata is transmitted
</mje>
> c) The filters on a consumer/channel may specify an filter based on
> event types. This is the one thing that is specified by our spec and
> will persist even if we adopt WSRA's EDL. The filter schema consists of
> two parts:
>
> eventType -- an abstract element (substitution group head)
> eventType.sca -- concrete element that uses our EDL (or will point to
> WSRA's EDL, if adopted) and is part of the substitution group. This
> requires that the definition of the event type must use XML schema as
> defined by our spec (currently) or WS-RA, if adopted.
>
> One can extend this by creating eventType.java or
> eventType.myFavoriteExtension
>
> It is important to note that the eventTypeXSD syntactic element defined
> by the EDL (in (a) above), cannot be used for filtering. The filtering
> element points to one or more QNames/namespaces used by the EDL. The
> structure of the elements is very different. This is the main error
> pointed out by issue 238 wrt 1.2-wd01 that this proposal fixes.
>
> An example:
>
> (a) EDL --
>
> <eventDefinitions targetNamespace="http://example.org/edl" ...>
> <eventTypeXSD name="PrinterEvent" element="foo:PrinterData"/>
> </eventDefinitions>
>
> (b) Event instance, if it used XML (made up example) --
>
> <foo:PrinterData ...>
> <foo:header>
> <foo:eventType>...</foo:eventType>
> ...
> </foo:header>
> <foo:body>
> ...
> </foo:body>
> </foo:PrinterData>
>
> (c) filter on the event type PrinterEvent defined by the EDL above --
>
> <filters xmlns:ex="http://example.org/edl" ...>
> <eventType.sca qnames="ex:PrinterEvent" />
> </filters>
>
>
> -Anish
> --
---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail. Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU