Andrew Zhang wrote:
On 7/31/06, Mikhail Fursov <[EMAIL PROTECTED]> wrote:
On 7/31/06, Andrew Zhang <[EMAIL PROTECTED]> wrote:
>
> Shall we follow RI or spec? I'd like to suggest follow RI for this
> specific
> case, because:
>
> 1. If we follow spec, it may cause legacy applications fail.
I do not think that fixing this bug will affect legacy applications
badly,
because we will add new functionality that is turned on by special
configuration files. Old functionality will not be affected.
Even with our special configuration, Harmony still behaves differently
from
RI, right?
For some case, RI could load user specified configuration file while
harmony
could not, and vice versa.
So I think legacy applications will be affected if they assume system
class
loader is used. :)
2. Other service provider in nio module are all loaded by system class
> loader.
>
> Any suggestions and comments?
This bug is 4 years old. One or two years more and it will become a
'feature'
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4619777
Anyway, I vote to follow spec but not RI :) (AFAIK as classpath did)
I tried to run the test against sun 6.0 pre-rc-b65. The result is the
same.
Seems SUN doesn't take any action on this problem. As you mentioned, it's
already 4 years, and is not solved in jdk6. :(
So I still think we should follow RI this time. :)
Shall we integrate the spec and RI? That is,
1) Try to load the Charset provider via the current thread's context
class loader
2) If fail to load the provider in step 1), try to use system class
loader to load the Charset provider.
Please correct me if I'm wrong. Thanks a lot.
Best regards,
Richard
Thanks!
--
Mikhail Fursov
--
Richard Liang
China Software Development Lab, IBM
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]