> -----Original Message-----
> From: Stefan Bodewig [mailto:[EMAIL PROTECTED]
> Sent: Thursday, 21 February 2002 1:10 AM
> To: [EMAIL PROTECTED]
> Subject: Re: antlib descriptor
>
>
> On Wed, 20 Feb 2002, Adam Murdoch <[EMAIL PROTECTED]> wrote:
>
> > I want to change the Foo.addBar() case to work similar to the way
> > Mutant does, with a 'type' attribute.  The type attribute would be
> > optional.
>
> [...]
>
> > Given that myrmidon doesn't actually use roles during configuration,
>
> In my previous mail I assumed, it did, sorry.
>
> Maybe it should, instead of using an additional attribute like Mutant?
> This could lead to an inflation of roles, though.
>

Yes, it is about to start using roles again.  Mainly because roles are a
handy place to attach meta-info to (e.g. things like the default
implementation, or specifying a particular type factory to use).  Once that
is done, I want to write up how configuration in myrmidon works, so people
can get a clearer picture of it (it *does* sound complicated in all these
emails, but it really isn't).

As far as inflation of roles goes, well, Myrmidon already has dozens of
them.  There is a role for pretty much everything that is pluggable.  And
there's a lot of those.


> > we should decide how syntax is going to work, and then worry about
> > whether roles should be a visible part of the solution.
>
> If we make roles invisible in the descriptor, we make it impossible
> for library providers to define their own pluggable elements in just
> that descriptor, don't we?
>

Well, yes.  For now, I think we should leave out the ability for an antlib
to define new roles (ie new groups of pluggable elements).  Let's stick with
'task' and 'data-type'.  Of course, that doesn't mean we won't be using the
roles concept internally.


Adam


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to