You are viewing a plain text version of this content. The canonical link for it is here.
Posted to server-dev@james.apache.org by Stefano Bagnara <ap...@bago.org> on 2006/11/08 12:51:58 UTC

Re: Next-major plans

Danny Angus wrote:
> On 11/7/06, Stefano Bagnara <ap...@bago.org> wrote:
> 
>> Reading what's going on on the mailet api side I think that we won't
>> include any change on mailet apis in next-major. The current proposal is
>> a strawman implementation and it needs a full non backward compatible
>> rewrite of the server.
> 
> I hope that over the next week or so I will be in a position to make
> some detailed proposals about the API. Andy Oliver has been great to
> bounce ideas off.

Gossip: I read Andy's blog, if I remember correctly he's no more the 
JBoss Mail Server guy.

> The fork is indeed a straw man, but please note that it isn't just the
> API thats being changed, I'm testing my ideas out by changing the
> server as well. In fact the head of the fork builds and runs with a
> fairly comprehensive array of mailets deployed. One of my aims is to
> ensure that James can be modified without too much pain to support the
> API changes.

I review every of your commit :-) You know I "hate" JNDI, but I think 
there are also good things in your commit.
- I like the MailFactory
- I like the string based lookup for repositories
- I don't like a rich "User" object in the API
- I like the DataSource change,
- I don't like the MailetContext => JNDI context move for simple value 
like HELLO_NAME and DEFAULT_DOMAIN (more code needed, less clear)

I have technical concern behind "likes" and "dislikes", but I think I 
said most of them, and I can give further details on demand.
Hope this is a useful and clear summary of my view.

> I'd also like to move the API out to its own sub-project before we
> change it, so that the API can be released with a test-bed of some
> kind to allow mailet authors to change their mailets before they
> upgrade the server, and to allow Server to upgrade to support the API
> changes at its own pace, and not forced by the pace of change of the
> API.

I wrote a bunch of times that I guess your Mailet APIs proposal do not 
belong to next-major (backward compatible / branch in 1-2 months) 
release, but more probably to the following one.
Can you confirm this? If so, please record this in the STATUS ;-)

>> Can you write what you think "we" have decided as I understand you don't
>> feel you have decided the following:
>> ------------
>> "next-major":
>> - based on current trunk
>> - storage and config.xml compatible with 2.3.0
>> - ETA: branch on Dec 2006/Jan 2007, release on Mar 2007
>> ---------
> 
> Please people, if we think we've agreed something record this stuff in
> the STATUS file then we can stop discussing it and move forwards. :-)
> 
> d.

When you wrote this I already updated that file including this data: the 
question now is "do we agree that we've agreed" ? ;-)
I thought we agreed but Noel for sure "ignore" this, and Bernd raised 
some concern on the hardness of the dates: do we need another 
discussion? Do we need another vote to really understand what we want to 
do? I'm lost, please help.

Stefano


---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org


Re: Next-major plans

Posted by Danny Angus <da...@gmail.com>.
> Gossip: I read Andy's blog, if I remember correctly he's no more the
> JBoss Mail Server guy.

Thats right, but I've known Andy for a while and we don't always
agree, so getting consensus between the two of us is an indication
that the proposals might satisfy both James and non-James
applications.

> I review every of your commit :-) You know I "hate" JNDI, but I think
> there are also good things in your commit.
> - I like the MailFactory

Thank you, I've meant to do that since *forever* but never got round to it.

> - I like the string based lookup for repositories

looking up repositories is one thing which we do over and over again,
so I figure it is a good candidate for making it easy. What we need to
do is to ensure that it continues to apply to the masilbox manager
strategy, and that the API doesn't need to contain any knowledge of
James naming strategy. At most just mailet config params

> - I don't like a rich "User" object in the API

No neither do I! I'm working on a completely new model of
users/accounts/domains/addresses to replace it with but I'm going
round in circles.

> - I like the DataSource change,

Yeah, we need to support drivers which already offer
javax.sqlDataSource implementations (e.g Oracle)

> - I don't like the MailetContext => JNDI context move for simple value
> like HELLO_NAME and DEFAULT_DOMAIN (more code needed, less clear)

Yeah, I'm not too happy with it either but I'm not decided about what
would be good, we need to ensure that we don't prescribe (force on
people) a closed set of parameters.
Perhaps "Properties MailetContext.getProperties()" might be the answer.

>
> I have technical concern behind "likes" and "dislikes", but I think I
> said most of them, and I can give further details on demand.
> Hope this is a useful and clear summary of my view.

Yes great.


> I wrote a bunch of times that I guess your Mailet APIs proposal do not
> belong to next-major (backward compatible / branch in 1-2 months)
> release, but more probably to the following one.
> Can you confirm this? If so, please record this in the STATUS ;-)

At the moment there is no decision on what the changes will be, when
there is we can talk about timing.



> When you wrote this I already updated that file including this data:

Yes I saw that, thank you and well done.

> the
> question now is "do we agree that we've agreed" ? ;-)
> I thought we agreed but Noel for sure "ignore" this, and Bernd raised
> some concern on the hardness of the dates: do we need another
> discussion? Do we need another vote to really understand what we want to
> do? I'm lost, please help.

I think that if you have called a vote and recorded the result then we
have agreed.
If there is still doubt or disagreement I suggest we discuss it in a
separate thread on general@ because that is a PMC matter.
However I think that it might help if we could formulate some simple
VOTEs to clarify each point separately. I'll look into this.

d.

Re: Next-major plans

Posted by Danny Angus <da...@gmail.com>.
> Gossip: I read Andy's blog, if I remember correctly he's no more the
> JBoss Mail Server guy.

Thats right, but I've known Andy for a while and we don't always
agree, so getting consensus between the two of us is an indication
that the proposals might satisfy both James and non-James
applications.

> I review every of your commit :-) You know I "hate" JNDI, but I think
> there are also good things in your commit.
> - I like the MailFactory

Thank you, I've meant to do that since *forever* but never got round to it.

> - I like the string based lookup for repositories

looking up repositories is one thing which we do over and over again,
so I figure it is a good candidate for making it easy. What we need to
do is to ensure that it continues to apply to the masilbox manager
strategy, and that the API doesn't need to contain any knowledge of
James naming strategy. At most just mailet config params

> - I don't like a rich "User" object in the API

No neither do I! I'm working on a completely new model of
users/accounts/domains/addresses to replace it with but I'm going
round in circles.

> - I like the DataSource change,

Yeah, we need to support drivers which already offer
javax.sqlDataSource implementations (e.g Oracle)

> - I don't like the MailetContext => JNDI context move for simple value
> like HELLO_NAME and DEFAULT_DOMAIN (more code needed, less clear)

Yeah, I'm not too happy with it either but I'm not decided about what
would be good, we need to ensure that we don't prescribe (force on
people) a closed set of parameters.
Perhaps "Properties MailetContext.getProperties()" might be the answer.

>
> I have technical concern behind "likes" and "dislikes", but I think I
> said most of them, and I can give further details on demand.
> Hope this is a useful and clear summary of my view.

Yes great.


> I wrote a bunch of times that I guess your Mailet APIs proposal do not
> belong to next-major (backward compatible / branch in 1-2 months)
> release, but more probably to the following one.
> Can you confirm this? If so, please record this in the STATUS ;-)

At the moment there is no decision on what the changes will be, when
there is we can talk about timing.



> When you wrote this I already updated that file including this data:

Yes I saw that, thank you and well done.

> the
> question now is "do we agree that we've agreed" ? ;-)
> I thought we agreed but Noel for sure "ignore" this, and Bernd raised
> some concern on the hardness of the dates: do we need another
> discussion? Do we need another vote to really understand what we want to
> do? I'm lost, please help.

I think that if you have called a vote and recorded the result then we
have agreed.
If there is still doubt or disagreement I suggest we discuss it in a
separate thread on general@ because that is a PMC matter.
However I think that it might help if we could formulate some simple
VOTEs to clarify each point separately. I'll look into this.

d.

---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org