[RFC][PATCH][icedtea-web]: Added support for signed JNLP file- Updated Patch

[Omair Majid]

On 07/21/2011 09:31 AM, Dr Andrew John Hughes wrote:
> On 09:21 Thu 21 Jul     , Jiri Vanek wrote:
>>>
>>> Can xml not return the number of element names beforehand?
>> Bad luck. No:(
>>>
>>> I'm not familiar with this XML API.  Where is it from?  Doesn't look like JAXP.
>> This is tiny html parser coded directly inside icedtea-web.... not sure if it is right choice ... but is is here very long tim and a several things depnde on him. It was chosen because he is much m,ore tolerant then xml parsers.
>> Weather it is benefit is question,, but it was decided long ago and is working fine.
>>
>

IcedTea-Web (or Netx, if you will, since this part of code has not been 
significantly modified) uses a (quite old) embedded copy of the NanoXML 
parser [1]. It had a number of problems (it could not even deal with 
comments when we first added netx to icedtea), but I think we have 
reached a point where it works fine for most users.

> There was a plan to replace that.  Omair?
>

There still is.

First, let me clarify something about jnlp files. They may look like 
(and they are defined to be, as far as I know) XML files. But thanks to 
the fact that there is just one reference implementation and it does not 
rely on a validating XML parser (as far as I can tell), the JNLP files 
are often not valid XML files. This causes a huge risk of regressions if 
we replace a 'XML' parser with another one.

That said, I did spend some time looking into replacing NanoXML with a 
JAXP-based SAX parser (actually, most of the code was in Netx but was 
just commented out). The unit tests pointed out a few regression, so I 
did not commit it. As much as I would love to remove the embedded 
NanoXML code, I am concerned about breaking applications.

Going forward, the plan is to switch to a parser designed for parsing 
malformed xml files. I have tested tagsoup and it looks somewhat 
promising [2]. But I am not very eager to do that in a minor release as 
the risk of regressions is still significant. It is currently scheduled 
for the 2.0 release [3], where I expect a few other major changes too.

Cheers,
Omair


[1] http://sourceforge.net/projects/nanoxml/
[2] 
http://mail.openjdk.java.net/pipermail/distro-pkg-dev/2011-January/011681.html
[3] http://icedtea.classpath.org/wiki/IcedTea-Web#IcedTea-Web_2.0