BUG: buffering prevents incremental parsing

[David Waite <ma...@ufl.edu>]

Setup:
Sun JDK 1.2.2 for Linux (also tested against 1.3 beta for Linux)
using the org.apache.xerces.parsers.SAXParser
based on current Xerces CVS (checked up to this evening)

Using my SAX code for the Jabber project (http://www.jabber.org), I have
been unable to get the SAX parser contained in Xerces to correctly parse
the incoming data incrementally. This code has been checked against an
alternate parser (AElfred) and found to work.

The core transport layer is XMLStreams, described at
http://etherx.jabber.org/streams/ and soon to be an Internet Draft.
Summerized as "Open socket, stuff XML".

There is a bug, which I believe to be a buffering problem, which is
causing messages to be detained until sufficient data arrives to push
the message out. My current estimates put this at about 16
characters/bytes. This is after modifying the Factory to return
StreamingCharReaders, as described in the Xerces FAQ.

Currently I have switched over to AElfred (which works very well, and is
significantly smaller than Xerces). I would however like to have this
issue resolved, so that I can support Xerces and xml4j as well for any
java applications which are already using these parsers, and would like
to also use my code.

I have posted two comments on this issue in as many days without
receiving any replies; I haven't even received flames. I have spent over
a week trying to track this problem down and failed, so without feedback
this will be the last time I remark on this issue in the forseeable
future. However, I *really do* want to support the Xerces parser, and am
willing to help out in tracking down and fixing this issue.

-David Waite

Re: buffering prevents incremental parsing

[Ted Leung <tw...@sauria.com>]

Hi David,

I originally did the modifications to allow the method in the FAQ to work.
Unfortunately, I am no longer working on Xerces as
my full time job.  I am willing to take a look at this for you, but it will
probably
take some time - around a week for me to get to it.

One thing you can do in the mean time is write a small test case:
one class sits on a socket and reads from the stream, and prints out
characters as they appear
another class writes a file to a socket in variable size chunks.

I had test code like this, but it stayed at IBM after I left.

It will also help me help you if you contact me via my new work e-mail
address, as I'm not able to read xerces-dev every day.

ted@enkubator.com

Ted
----- Original Message -----
From: "David Waite" <ma...@ufl.edu>
To: <xe...@xml.apache.org>
Sent: Wednesday, June 14, 2000 11:08 PM
Subject: BUG: buffering prevents incremental parsing


> Setup:
> Sun JDK 1.2.2 for Linux (also tested against 1.3 beta for Linux)
> using the org.apache.xerces.parsers.SAXParser
> based on current Xerces CVS (checked up to this evening)
>
> Using my SAX code for the Jabber project (http://www.jabber.org), I have
> been unable to get the SAX parser contained in Xerces to correctly parse
> the incoming data incrementally. This code has been checked against an
> alternate parser (AElfred) and found to work.
>
> The core transport layer is XMLStreams, described at
> http://etherx.jabber.org/streams/ and soon to be an Internet Draft.
> Summerized as "Open socket, stuff XML".
>
> There is a bug, which I believe to be a buffering problem, which is
> causing messages to be detained until sufficient data arrives to push
> the message out. My current estimates put this at about 16
> characters/bytes. This is after modifying the Factory to return
> StreamingCharReaders, as described in the Xerces FAQ.
>
> Currently I have switched over to AElfred (which works very well, and is
> significantly smaller than Xerces). I would however like to have this
> issue resolved, so that I can support Xerces and xml4j as well for any
> java applications which are already using these parsers, and would like
> to also use my code.
>
> I have posted two comments on this issue in as many days without
> receiving any replies; I haven't even received flames. I have spent over
> a week trying to track this problem down and failed, so without feedback
> this will be the last time I remark on this issue in the forseeable
> future. However, I *really do* want to support the Xerces parser, and am
> willing to help out in tracking down and fixing this issue.
>
> -David Waite
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: xerces-j-dev-unsubscribe@xml.apache.org
> For additional commands, e-mail: xerces-j-dev-help@xml.apache.org
>
>
>