[Devel] lv2_descriptor() considered harmful

[Stefano D'Angelo]

2011/10/10 David Robillard <d at drobilla.net>:
> On Mon, 2011-10-10 at 11:08 +0300, Stefano D'Angelo wrote:
>> 2011/10/10 David Robillard <d at drobilla.net>:
>> >  * Missing bundle_path parameter, making more "dynamic" plugin bundles
>> > effectively impossible (e.g. any bundle that needs to run some code at
>> > init time to see what plugins it can make available, such as plugins
>> > defined in some data file.  Think plugins in Faust or Ingen patches)
>>
>> Maybe I don't understand correctly, but maybe there is some overlap
>> with dynamic manifest here? In other words, you still want all the
>> "potential plugins" to be listed in manifest.ttl or not?
>
> My immediate use case is for Ingen, where the description of the
> plugin/patch is static, but the library needs to be able to read that
> data to know what plugin it is providing.  It reads the data to know
> what ports are there, what the plugin URI is, etc. etc.
>
> I could perhaps solve this problem with dynamic manifest, but that is an
> overly complicated and slow solution here, since I don't actually need
> to dynamically generate any RDF data at all.

Ah, ok.

> In short, I see lv2_descriptor as flawed since it makes it impossible to
> implement plugins are defined by files in their bundle (essentially any
> plugin that is not statically written and compiled C code).  Faust is
> another good example of such a thing.

Agreed (except for FAUST, I don't believe that'd be a clever strategy
- just generate the necessary C code and RDF data, but whatever).

Stefano

P.S.: I was not asking because I'm attached to dynmanifest in any way,
I was doing that because of the FAUST-like thing I am going to do. ;-)