Yes. But code should NOT rely that it will get CRLFs from the client. For
instance some of perl libraries, lib-www and MIME::Tools, are breaking the
spec and are sending just LFs.
-Original Message-
From: Tom Jordahl [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, June 04, 2002 11:20 AM
To: '[E
CRLF sequences are required by rfc 822 for the headers. They are also used
in MIME as delimiters. In some instances, rfc 2049, it is acceptable to use
system's local representation of new line, but CRLF is preferred. Some
parser out there may understand single LF or just CR, but all are going to
u
in some instances headers were written followed by only LineFeed, when CRLF
is expected.
RCS file:
/home/cvspublic/xml-axis/java/src/org/apache/axis/transport/http/HTTPSender.
java,v
retrieving revision 1.53
diff -w -r1.53 HTTPSender.java
169c169
< out.print("CONNECT " + h
I have noticed a small problem with the Deserializer.
Axis is unable to resolve namespace mapping if it has been defined on the
same level.
The following tag will not have its type picked up because namespace mapping
will fail.
***
http://my.org/schemas/service";
xsi:type="ens:emailScript">
***
I
>I have folded the function of your patch into my sandbox, but I wonder
>if you could provide the rationale (e.g. what bug did it fix?).
There was no way to construct XML for the headers by hand. Also in the body
you had to use RPCElement to create nodes.
As well as a fix for serialization of chil
This is the patch we were talking about in chat.
set/getValue changed to set/getObjectValue()
-Taras
SOAPHeader.java
Description: Binary data
MessageElement.java
Description: Binary data
Constants.diff
Description: Binary data
One of the MAJOR issues that I have noted with axis is the way it parses the
data. In a sense you data is parsed in 2 passes.
Originally what happens is Axis deserializes XML into MessageElement objects
constructing a DOM structure. At this time all of the SAX events are
recorded.
When that is don
c: '[EMAIL PROTECTED]'
> Subject: RE: Experiments on type serialization. (feedback
> very welcome)
>
>
> Taras,
>
> Could you resend this patch to me.
>
> Thanks
>
> Rich Scheuerle
> XML & Web Services Development
> 512-838-5115 (IBM TL 678-5115)
>
>
>
o examine the ramifications of pooling some of the
> objects versus moving logical function, that would
> be a great help to the group.
>
> Thanks,
>
> Rich Scheuerle
> XML & Web Services Development
> 512-838-5115 (IBM TL 678-5115)
>
>
>
>
= null) {
< super.outputImpl(context);
< } else {
< context.serialize(new QName(namespaceURI,name), attributes,
value, value.getClass() );
< }
< }
Constants***
382a383
> public static final String ATTR_TYPE
ch Scheuerle
> XML & Web Services Development
> 512-838-5115 (IBM TL 678-5115)
>
>
>
>
> Taras Shkvarchuk
>
If node had an href attribute, but that reference was not defined yet, you
would get NullPointer in log4j DEBUG mode.
316a317
> if(ref!=null){
319a321,325
> }else{
> category.debug(JavaUtils.getMessa
org.apache.axis.message.MessageElement:
added Object getValue()/setValue(Object) methods.
makes it a lot easier to build nodes with registered types as values.
Fixed output() method to serialize any children node may have.
-Taras
61a62
> import org.apache.axis.encoding.Serializer;
80a82
> impo
815c815
< return null;
---
> return dser;
-- Lazy (on demand) parsing is definitely an area for improvement.
In intermediary services much of the data in the message is not used.
-- Particularly in Axis, direct serialization of XML into registered types
(using SAX of cause). In some instances that gives as much as 1/3 memory
utilization
Ok this change is actually working right. Having attachment support
optional makes it not auto rebuild on dependency checks :(
> -Original Message-
> From: Taras Shkvarchuk
> Sent: Tuesday, February 12, 2002 1:51 PM
> To: '[EMAIL PROTECTED]'
> Subject: RE: Att
oops forgot to attach the change. :)
> -Original Message-
> From: Taras Shkvarchuk
> Sent: Tuesday, February 12, 2002 1:18 PM
> To: '[EMAIL PROTECTED]'
> Subject: RE: Attachment support Diffs for review (Message & Part &
> SOAPPar t af fected)
>
attachments
> from the incoming
> stream unless they are reference and thus
> I'm not going to integrate this change in.
>
>
> Rick Rineholt
> "The truth is out there... All you need is a better search engine!"
>
> [EMAIL PROTECTED]
>
>
&
Even though Axis is using SAX internally, it still creates a DOM like
structure from MessageElement object. Also all of the events are recorded.
This leads to cpu/memory utilization similar to that of DOM model. When an
object is deserialized from the Envelope, it is getting all of the events
from
Totally understand, specially since there is some design change.
Hope you find these changes useful. :)
Thanks!
> -Original Message-
> From: Rick Rineholt [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, February 06, 2002 8:07 AM
> To: [EMAIL PROTECTED]
> Subject: Re: Attachment support Dif
Here are some changes I made to Attachments implementation. (diffs attached)
*
AttachmentsImpl.java:
removed reference to Message object.
When Message gets SOAPPart it sets itself as a parent.
added orderedAttachments list that keep
> There may be some reasons for this to follow
> the JAXM model. But to make really sure I would need to
> do some more thorough investigation. In doing so what have you fixed?
Not having to depend on Message, you can share Attachments between messages.
In catalog type services you might want to
Currently SOAPBody.getBytes() just calls getBytes() on the string object.
This will use VM's default character set. That happens to be "Cp1252" on
most systems, so any non LATIN characters in will get corrupt.
Likewise, when creating a String from the raw bytes default encoding is
used.
I think
Attachment data stored in ManagedMemoryDataSource is deleted (or at least
marked deleted), as soon as message has been serialized.
While in most cases you will not want to access the data after you have sent
out the message, why force a delete possibly too early?
In case where a service is trying
I have been playing with Attachment code trying to improve it a little, and
fill in the missing parts/TODOs.
Doing that I am not understanding 2 basic things in the design.
1) Why would you want to compile Axis without attachment support? Doing so
creates, in my opinion, unneeded code complexity
25 matches
Mail list logo
